2026-04-16 00:58:36.493390: 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)
I0416 00:58:36.609224 133674723441792 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag.
I0416 00:59:06.525939 133674723441792 max_utils.py:800] System Information: Jax Version: 0.9.2
I0416 00:59:06.526087 133674723441792 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0416 00:59:06.526130 133674723441792 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Apr 6 2026 20:48:10 (1775533690) cl/895581894
I0416 00:59:06.526156 133674723441792 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system
I0416 00:59:06.526266 133674723441792 train.py:812] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper.
W0416 00:59:06.621796 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 00:59:07.028675 133674723441792 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0416 00:59:07.028866 133674723441792 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0416 00:59:07.029120 133674723441792 checkpointing.py:688] Setting up checkpoint logger...
I0416 00:59:07.029163 133674723441792 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0416 00:59:07.029201 133674723441792 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0416 00:59:07.030029 133674723441792 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 0x7992fc1f8cb0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0416 00:59:09.368865 133674723441792 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0416 00:59:09.369075 133674723441792 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 0x798c807910a0>}, handler_registry=None
I0416 00:59:09.369327 133674723441792 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x798c807910a0>` 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`.
I0416 00:59:09.369369 133674723441792 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x798c7f935400>` 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`.
I0416 00:59:09.369398 133674723441792 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 0x798c807910a0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x798c807910a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x798c7f935400>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x798c7f935400>}).
I0416 00:59:09.369708 133674723441792 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
I0416 00:59:09.369766 133674723441792 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x798c7f9420c0> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0416 00:59:09.491244 133674723441792 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints
I0416 00:59:09.491472 133674723441792 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_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x798c807638c0>
I0416 00:59:09.491550 133674723441792 checkpointing.py:302] Checkpoint manager created!
I0416 00:59:09.571868 133674723441792 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0
I0416 00:59:09.574898 133674723441792 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.
I0416 00:59:09.628915 133674723441792 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
I0416 00:59:09.664510 133674723441792 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2
I0416 00:59:09.664636 133674723441792 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2
I0416 00:59:10.842147 133674723441792 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0416 00:59:10.842281 133674723441792 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0416 00:59:10.945283 133674723441792 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0416 00:59:10.945368 133674723441792 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0416 00:59:10.960217 133674723441792 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0416 00:59:10.960268 133674723441792 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0416 00:59:11.006444 133674723441792 attentions.py:1154] attentions/query Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0416 00:59:11.006518 133674723441792 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0416 00:59:11.021529 133674723441792 attentions.py:1155] attentions/key Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0416 00:59:11.021584 133674723441792 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0416 00:59:11.036483 133674723441792 attentions.py:1156] attentions/value Logical: bfloat16[8,2048,16,128]..................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim').
I0416 00:59:11.036538 133674723441792 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0416 00:59:11.058751 133674723441792 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0416 00:59:11.058809 133674723441792 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0416 00:59:11.103681 133674723441792 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0416 00:59:11.103749 133674723441792 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None).
I0416 00:59:11.486547 133674723441792 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0416 00:59:11.486652 133674723441792 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
W0416 00:59:11.554753 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
[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
I0416 00:59:12.330395 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330501 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330539 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330571 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330596 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330620 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330643 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330665 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330687 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330710 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330738 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330760 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330781 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330803 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330823 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330843 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330862 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330882 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330901 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330921 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330940 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.330960 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.330979 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331000 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331020 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331039 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331076 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331097 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331117 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331136 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331157 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331177 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331197 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331216 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331236 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331256 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331276 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331296 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331316 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331336 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331355 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 00:59:12.331374 133674723441792 maxtext_utils.py:1790] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 00:59:12.331413 133674723441792 maxtext_utils.py:1790] params/params/decoder/decoder_norm/scale
Shape: float32[2048]
Logical: P('norm',)
Physical: (None,)
I0416 00:59:12.331454 133674723441792 maxtext_utils.py:1790] params/params/decoder/layers/mlp/wi_0/kernel
Shape: float32[2048,16,7168]
Logical: P('embed', 'layers', 'mlp')
Physical: ('fsdp', None, None)
I0416 00:59:12.331480 133674723441792 maxtext_utils.py:1790] params/params/decoder/layers/mlp/wi_1/kernel
Shape: float32[2048,16,7168]
Logical: P('embed', 'layers', 'mlp')
Physical: ('fsdp', None, None)
I0416 00:59:12.331515 133674723441792 maxtext_utils.py:1790] params/params/decoder/layers/mlp/wo/kernel
Shape: float32[7168,16,2048]
Logical: P('mlp', 'layers', 'embed')
Physical: (None, None, 'fsdp')
I0416 00:59:12.331548 133674723441792 maxtext_utils.py:1790] params/params/decoder/layers/post_self_attention_layer_norm/scale
Shape: float32[2048,16]
Logical: P('norm', 'layers')
Physical: (None, None)
I0416 00:59:12.331572 133674723441792 maxtext_utils.py:1790] params/params/decoder/layers/pre_self_attention_layer_norm/scale
Shape: float32[2048,16]
Logical: P('norm', 'layers')
Physical: (None, None)
I0416 00:59:12.331607 133674723441792 maxtext_utils.py:1790] 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)
I0416 00:59:12.331641 133674723441792 maxtext_utils.py:1790] 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')
I0416 00:59:12.331667 133674723441792 maxtext_utils.py:1790] 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)
I0416 00:59:12.331692 133674723441792 maxtext_utils.py:1790] 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)
I0416 00:59:12.331729 133674723441792 maxtext_utils.py:1790] params/params/decoder/logits_dense/kernel
Shape: float32[2048,32000]
Logical: P('embed_vocab', 'vocab')
Physical: ('fsdp', None)
I0416 00:59:12.331762 133674723441792 maxtext_utils.py:1790] params/params/token_embedder/embedding
Shape: float32[32000,2048]
Logical: P('vocab', 'embed_vocab')
Physical: (None, 'fsdp')
I0416 00:59:13.760627 133674723441792 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0416 00:59:13.760725 133674723441792 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0416 00:59:13.775381 133674723441792 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0416 00:59:13.775434 133674723441792 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
W0416 00:59:14.257922 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 00:59:14.442420 133674723441792 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.
I0416 00:59:14.442990 133674723441792 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter.
I0416 00:59:14.443230 133674723441792 metric_logger.py:289] number parameters: 1.105 billion
W0416 00:59:18.760922 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 00:59:18.975201 133674723441792 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0416 00:59:19.089001 133674723441792 checkpointing.py:798] Waited 0.11377763748168945 seconds for step 0 to finish before starting checkpointing.
I0416 00:59:19.089517 133674723441792 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0416 00:59:19.089672 133674723441792 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0
I0416 00:59:19.090098 133674723441792 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0.
I0416 00:59:19.174256 133674723441792 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient
I0416 00:59:19.258870 133561342494272 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0
I0416 00:59:19.268360 133674723441792 jax_array_handlers.py:347] Scheduling D2H of 81 prioritized jax.Array.
I0416 00:59:19.268459 133674723441792 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
I0416 00:59:19.960086 133561332008512 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0/items
W0416 00:59:19.991367 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 00:59:20.005422 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 00:59:20.018328 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 00:59:20.032655 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 00:59:20.041881 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 00:59:20.048025 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 00:59:20.052918 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 00:59:20.058164 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 00:59:20.112912 133561292162624 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776301159868903520, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776301159868903520, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0/_CHECKPOINT_METADATA
I0416 00:59:20.505822 3458896 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json
I0416 00:59:20.505880 3458896 google_auth_provider.cc:156] Using OAuth2 AuthProvider
I0416 00:59:20.507907 133674723441792 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 1.240092s
I0416 00:59:20.515495 133674723441792 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 9.213 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host)
I0416 00:59:20.515563 133674723441792 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.339748s (batch_requests_ready=0.085734s, total_serialization_initiated=1.246569s, others=0.007444s)
I0416 00:59:20.515640 133674723441792 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.340861s (all_items=0.000021s, per_item={'items': '0.00002074'}, temp_paths=1.340841)
I0416 00:59:20.516855 133561208276544 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0416 00:59:20.516962 133674723441792 async_checkpointer.py:561] Finished blocking save. Time taken: 1.427248s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0.
I0416 00:59:20.517185 133674723441792 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0416 00:59:20.517365 133561352980032 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0416 00:59:20.517464 133674723441792 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776301159.0895047, 'wait_for_prev_duration_secs': 4.124641418457031e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776301159.0896935, 'checkpointer_blocking_duration_secs': 1.4274134635925293, 'get_old_steps_start_time': 1776301160.5171266, 'get_old_steps_duration_secs': 2.288818359375e-05, 'checkpoint_manager_blocking_start_time': 1776301159.0894165, 'checkpoint_manager_blocking_duration_secs': 1.4280197620391846}
I0416 00:59:20.517578 133674723441792 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0416 00:59:20.517633 133674723441792 max_utils.py:750]
Memstats: After params initialized:
I0416 00:59:20.517681 133674723441792 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_0(process=0,(0,0,0,0))
I0416 00:59:20.517707 133674723441792 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_1(process=0,(1,0,0,0))
I0416 00:59:20.517726 133674723441792 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_2(process=0,(0,1,0,0))
I0416 00:59:20.517745 133674723441792 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_3(process=0,(1,1,0,0))
I0416 00:59:20.517764 133674723441792 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_4(process=0,(0,2,0,0))
I0416 00:59:20.517781 133674723441792 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_5(process=0,(1,2,0,0))
I0416 00:59:20.517798 133674723441792 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_6(process=0,(0,3,0,0))
I0416 00:59:20.517814 133674723441792 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_7(process=0,(1,3,0,0))
W0416 00:59:20.524412 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 00:59:20.530462 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 00:59:20.535165 3456220 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 00:59:20.875085 133674723441792 metric_logger.py:185] completed step: 0, seconds: 4.531, TFLOP/s/device: 2.998, Tokens/s/device: 451.949, total_weights: 13328, loss: 10.844
I0416 00:59:20.876033 133674723441792 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/tensorboard/'
I0416 00:59:20.983014 133674723441792 metric_logger.py:185] completed step: 1, seconds: 1.897, TFLOP/s/device: 7.162, Tokens/s/device: 1079.582, total_weights: 12332, loss: 10.871
I0416 00:59:21.093145 133674723441792 metric_logger.py:185] completed step: 2, seconds: 0.011, TFLOP/s/device: 1271.608, Tokens/s/device: 191670.566, total_weights: 15161, loss: 10.871
I0416 00:59:21.203413 133674723441792 metric_logger.py:185] completed step: 3, seconds: 0.109, TFLOP/s/device: 124.415, Tokens/s/device: 18753.205, total_weights: 13327, loss: 10.857
I0416 00:59:21.494703 133561250219584 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0/items/array_metadatas/process_0
I0416 00:59:44.386548 133674723441792 metric_logger.py:185] completed step: 4, seconds: 0.110, TFLOP/s/device: 123.242, Tokens/s/device: 18576.301, total_weights: 11939, loss: 10.861
I0416 00:59:44.395664 133674723441792 metric_logger.py:185] completed step: 5, seconds: 0.110, TFLOP/s/device: 123.302, Tokens/s/device: 18585.404, total_weights: 15502, loss: 10.882
I0416 00:59:44.504859 133674723441792 metric_logger.py:185] completed step: 6, seconds: 23.182, TFLOP/s/device: 0.586, Tokens/s/device: 88.345, total_weights: 13864, loss: 10.886
I0416 00:59:44.614953 133674723441792 metric_logger.py:185] completed step: 7, seconds: 0.007, TFLOP/s/device: 2080.724, Tokens/s/device: 313629.403, total_weights: 12988, loss: 10.829
I0416 00:59:44.688657 133561239733824 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 24.172423s (commit=23.742357s, array_metadata_write=0.430065s)
I0416 00:59:44.689894 133561208276544 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 495.340 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 25 seconds) (per-host)
I0416 00:59:44.690023 133561208276544 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 24.173010s.
I0416 00:59:44.725463 133674723441792 metric_logger.py:185] completed step: 8, seconds: 0.110, TFLOP/s/device: 123.029, Tokens/s/device: 18544.342, total_weights: 13820, loss: 10.848
I0416 00:59:44.732455 133674723441792 checkpointing.py:794] Waiting for step 10 to finish before checkpoint...
I0416 00:59:44.919108 133561208276544 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776301159868903520, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0/_CHECKPOINT_METADATA
I0416 00:59:44.945967 133674723441792 checkpointing.py:798] Waited 0.2134711742401123 seconds for step 10 to finish before starting checkpointing.
I0416 00:59:44.946456 133674723441792 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0416 00:59:45.094897 133561208276544 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0416 00:59:45.310300 133561292162624 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776301159868903520, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0/_CHECKPOINT_METADATA
I0416 00:59:45.522637 133561208276544 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0416 00:59:45.523282 133561208276544 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.563020s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0/items
I0416 00:59:45.524061 133561208276544 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0/items
I0416 00:59:45.762117 133561208276544 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0
I0416 00:59:46.443351 133561208276544 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0`.
I0416 00:59:46.444129 133561208276544 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 27.354424s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/0
I0416 00:59:46.444203 133561208276544 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 25.927191s.
I0416 00:59:46.444435 133561352980032 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0416 00:59:46.444561 133561352980032 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0416 00:59:46.444614 133561352980032 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0416 00:59:46.444657 133561352980032 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0416 00:59:46.444826 133674723441792 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.
W0416 00:59:46.444964 133674723441792 checkpoint_manager.py:1441] Waiting for previous save to complete took 1.498510 seconds. If this number is high, consider checkpointing less frequently.
I0416 00:59:46.445530 133674723441792 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10
I0416 00:59:46.445848 133674723441792 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10.
I0416 00:59:46.622952 133674723441792 jax_array_handlers.py:347] Scheduling D2H of 81 prioritized jax.Array.
I0416 00:59:46.623063 133674723441792 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
I0416 00:59:46.625790 133561352980032 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10
I0416 00:59:47.337213 133561197790784 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10/items
I0416 00:59:58.440506 133674723441792 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 11.818039s
I0416 00:59:58.441165 133674723441792 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.036 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 11 seconds) (per-host)
I0416 00:59:58.441232 133674723441792 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 11.909864s (batch_requests_ready=0.085152s, total_serialization_initiated=11.824182s, others=0.000530s)
I0416 00:59:58.441304 133674723441792 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 11.910422s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=11.910405)
I0416 00:59:58.449916 133561239733824 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0416 00:59:58.450031 133674723441792 async_checkpointer.py:561] Finished blocking save. Time taken: 12.004450s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10.
I0416 00:59:58.450276 133674723441792 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0416 00:59:58.450511 133561208276544 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0416 00:59:58.450632 133674723441792 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776301184.9464304, 'wait_for_prev_duration_secs': 1.4985098838806152, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776301186.4455547, 'checkpointer_blocking_duration_secs': 12.004637718200684, 'get_old_steps_start_time': 1776301198.450214, 'get_old_steps_duration_secs': 2.384185791015625e-05, 'checkpoint_manager_blocking_start_time': 1776301184.9463627, 'checkpoint_manager_blocking_duration_secs': 13.504240989685059}
I0416 00:59:58.450763 133674723441792 checkpointing.py:409] Started an asynchronous checkpoint save for step 10
I0416 00:59:58.451267 133674723441792 metric_logger.py:185] completed step: 9, seconds: 0.110, TFLOP/s/device: 123.325, Tokens/s/device: 18588.946, total_weights: 12300, loss: 10.857
I0416 00:59:58.461489 133674723441792 metric_logger.py:185] completed step: 10, seconds: 0.111, TFLOP/s/device: 122.937, Tokens/s/device: 18530.415, total_weights: 13436, loss: 10.876
I0416 00:59:58.568587 133674723441792 metric_logger.py:185] completed step: 11, seconds: 13.726, TFLOP/s/device: 0.990, Tokens/s/device: 149.203, total_weights: 13817, loss: 10.853
I0416 00:59:58.679215 133674723441792 metric_logger.py:185] completed step: 12, seconds: 0.008, TFLOP/s/device: 1670.412, Tokens/s/device: 251782.641, total_weights: 12076, loss: 10.852
I0416 00:59:58.789560 133674723441792 metric_logger.py:185] completed step: 13, seconds: 0.109, TFLOP/s/device: 124.711, Tokens/s/device: 18797.786, total_weights: 11048, loss: 10.864
I0416 00:59:58.899815 133674723441792 metric_logger.py:185] completed step: 14, seconds: 0.111, TFLOP/s/device: 122.794, Tokens/s/device: 18508.812, total_weights: 14267, loss: 10.834
I0416 00:59:59.010007 133674723441792 metric_logger.py:185] completed step: 15, seconds: 0.110, TFLOP/s/device: 123.303, Tokens/s/device: 18585.573, total_weights: 13004, loss: 10.832
I0416 00:59:59.120343 133674723441792 metric_logger.py:185] completed step: 16, seconds: 0.110, TFLOP/s/device: 123.343, Tokens/s/device: 18591.646, total_weights: 13811, loss: 10.867
I0416 00:59:59.178869 133561271191104 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10/items/array_metadatas/process_0
I0416 00:59:59.230825 133674723441792 metric_logger.py:185] completed step: 17, seconds: 0.110, TFLOP/s/device: 123.300, Tokens/s/device: 18585.067, total_weights: 13709, loss: 10.850
I0416 00:59:59.348968 133674723441792 metric_logger.py:185] completed step: 18, seconds: 0.111, TFLOP/s/device: 122.953, Tokens/s/device: 18532.763, total_weights: 12831, loss: 10.839
I0416 00:59:59.451530 133674723441792 checkpointing.py:794] Waiting for step 19 to finish before checkpoint...
I0416 00:59:59.453341 133674723441792 checkpointing.py:798] Waited 0.0018224716186523438 seconds for step 19 to finish before starting checkpointing.
I0416 00:59:59.453810 133674723441792 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0416 01:00:22.978734 133561197790784 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 24.529768s (commit=24.117013s, array_metadata_write=0.412755s)
I0416 01:00:22.979985 133561239733824 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 346.738 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 36 seconds) (per-host)
I0416 01:00:22.980107 133561239733824 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 24.530005s.
I0416 01:00:23.392245 133561239733824 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0416 01:00:23.802284 133561239733824 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0416 01:00:23.803037 133561239733824 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.551278s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10/items
I0416 01:00:23.803776 133561239733824 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10/items
I0416 01:00:24.040439 133561239733824 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10
I0416 01:00:24.741061 133561239733824 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10`.
I0416 01:00:24.741802 133561239733824 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 38.296234s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/10
I0416 01:00:24.741884 133561239733824 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 26.291784s.
I0416 01:00:24.742031 133561208276544 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0416 01:00:24.742157 133561208276544 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0416 01:00:24.742220 133561208276544 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0416 01:00:24.742262 133561208276544 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0416 01:00:24.742665 133674723441792 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.
W0416 01:00:24.742795 133674723441792 checkpoint_manager.py:1441] Waiting for previous save to complete took 25.289000 seconds. If this number is high, consider checkpointing less frequently.
I0416 01:00:24.743017 133674723441792 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19
I0416 01:00:24.743349 133674723441792 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19.
I0416 01:00:24.930764 133561208276544 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19
I0416 01:00:24.950766 133674723441792 jax_array_handlers.py:347] Scheduling D2H of 81 prioritized jax.Array.
I0416 01:00:24.950872 133674723441792 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
I0416 01:00:25.635015 133561260705344 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19/items
I0416 01:00:36.159429 133674723441792 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 11.209147s
I0416 01:00:36.167264 133674723441792 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.090 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 11 seconds) (per-host)
I0416 01:00:36.167374 133674723441792 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 11.318305s (batch_requests_ready=0.095584s, total_serialization_initiated=11.214957s, others=0.007764s)
I0416 01:00:36.167457 133674723441792 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 11.318884s (all_items=0.000013s, per_item={'items': '0.00001287'}, temp_paths=11.318871)
I0416 01:00:36.169059 133561271191104 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0416 01:00:36.169162 133674723441792 async_checkpointer.py:561] Finished blocking save. Time taken: 11.426091s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19.
I0416 01:00:36.169334 133674723441792 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0416 01:00:36.169529 133561218762304 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0416 01:00:36.169609 133674723441792 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776301199.4537814, 'wait_for_prev_duration_secs': 25.288999557495117, 'time_between_consecutive_saves_sec': 13.009109735488892, 'checkpointer_blocking_start_time': 1776301224.743041, 'checkpointer_blocking_duration_secs': 11.42623233795166, 'get_old_steps_start_time': 1776301236.169285, 'get_old_steps_duration_secs': 1.9550323486328125e-05, 'checkpoint_manager_blocking_start_time': 1776301199.4537337, 'checkpoint_manager_blocking_duration_secs': 36.71585011482239}
I0416 01:00:36.169695 133674723441792 checkpointing.py:409] Started an asynchronous checkpoint save for step 19
I0416 01:00:36.169725 133674723441792 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0416 01:00:36.902366 133561352980032 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19/items/array_metadatas/process_0
I0416 01:01:06.955692 133561260705344 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 30.787752s (commit=30.311474s, array_metadata_write=0.476278s)
I0416 01:01:06.956859 133561271191104 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 300.137 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 42 seconds) (per-host)
I0416 01:01:06.956981 133561271191104 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 30.787773s.
I0416 01:01:07.406346 133561271191104 array_metadata_store.py:367] [process=0][thread=async_save] Skipped cross-host ArrayMetadata validation because only one process is found: process_index=0.
I0416 01:01:07.838407 133561271191104 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0416 01:01:07.839183 133561271191104 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.583000s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19/items
I0416 01:01:07.839945 133561271191104 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19/items
I0416 01:01:08.104578 133561271191104 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19
I0416 01:01:08.799125 133561271191104 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19`.
I0416 01:01:08.799853 133561271191104 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 44.056788s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_trainstate_and_training_loop_20260416_004836/linen_feat_nnx_trainstate_and_training_loop_20260416_004836_05_fp8/checkpoints/19
I0416 01:01:08.799916 133561271191104 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 32.630710s.
I0416 01:01:08.800034 133561218762304 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0416 01:01:08.800096 133561218762304 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0416 01:01:08.800150 133561218762304 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0416 01:01:08.800185 133561218762304 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0416 01:01:08.800753 133674723441792 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.
I0416 01:01:08.800899 133674723441792 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0416 01:01:08.801582 133674723441792 metric_logger.py:185] completed step: 19, seconds: 0.111, TFLOP/s/device: 122.890, Tokens/s/device: 18523.376, total_weights: 13717, loss: 10.851
Per train step:
Total TFLOPs: 13.59
split as 93.93% learnable weight flops and 6.07% attention flops