2026-04-16 18:26:30.113464: 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 18:26:30.229005 123946223828096 max_utils.py:238] Skipping jax distributed system due to skip_jax_distributed_system=True flag.
I0416 18:27:16.589108 123946223828096 max_utils.py:800] System Information: Jax Version: 0.9.2
I0416 18:27:16.589224 123946223828096 max_utils.py:801] System Information: Jaxlib Version: 0.9.2
I0416 18:27:16.589264 123946223828096 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 18:27:16.589289 123946223828096 train_utils.py:364] WARNING: 'dataset_path' might be pointing your local file system
I0416 18:27:16.589377 123946223828096 train.py:811] [DECOUPLED NO-OP] skipping cloud diagnostics wrapper.
W0416 18:27:16.683902 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 18:27:17.087059 123946223828096 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0416 18:27:17.087232 123946223828096 maxtext_utils.py:1687] Num_devices: 8, shape (1, 1, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0416 18:27:17.087461 123946223828096 checkpointing.py:688] Setting up checkpoint logger...
I0416 18:27:17.087503 123946223828096 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True
I0416 18:27:17.087536 123946223828096 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0416 18:27:17.088203 123946223828096 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 0x70b9e45fce00>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0416 18:27:19.436668 123946223828096 checkpointing.py:266] Enabling policy for fixed interval checkpointing.
I0416 18:27:19.436838 123946223828096 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 0x70b36713a300>}, handler_registry=None
I0416 18:27:19.437095 123946223828096 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x70b36713a300>` 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 18:27:19.437137 123946223828096 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70b36713a330>` 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 18:27:19.437165 123946223828096 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 0x70b36713a300>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x70b36713a300>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70b36713a330>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x70b36713a330>}).
I0416 18:27:19.437445 123946223828096 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
I0416 18:27:19.437493 123946223828096 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>.<lambda> at 0x70b367131ee0> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0416 18:27:19.560981 123946223828096 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints
I0416 18:27:19.561229 123946223828096 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_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x70b367ddafc0>
I0416 18:27:19.561305 123946223828096 checkpointing.py:302] Checkpoint manager created!
I0416 18:27:19.640876 123946223828096 dataset_info.py:707] Load dataset info from tests/assets/local_datasets/c4_en_dataset_minimal/c4/en/3.1.0
I0416 18:27:19.643848 123946223828096 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 18:27:19.696947 123946223828096 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 18:27:19.724399 123946223828096 tokenizer.py:245] Tokenizer path: src/maxtext/assets/tokenizers/tokenizer.llama2
I0416 18:27:19.724462 123946223828096 tokenizer.py:187] Loading sentencepiece tokenizer: src/maxtext/assets/tokenizers/tokenizer.llama2
I0416 18:27:20.813896 123946223828096 nnx_wrappers.py:437] Unknown Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_norm_length', 'activation_embed').
I0416 18:27:20.813992 123946223828096 nnx_wrappers.py:437] Unknown Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0416 18:27:20.916926 123946223828096 attentions.py:1088] attentions/inputs_q Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0416 18:27:20.917005 123946223828096 attentions.py:1088] attentions/inputs_q Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0416 18:27:20.931863 123946223828096 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[8,2048,2048]....................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed').
I0416 18:27:20.931912 123946223828096 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[8,2048,2048]....................................... ('fsdp', None, None).
I0416 18:27:20.977944 123946223828096 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 18:27:20.978015 123946223828096 attentions.py:1154] attentions/query Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0416 18:27:20.992967 123946223828096 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 18:27:20.993018 123946223828096 attentions.py:1155] attentions/key Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0416 18:27:21.007910 123946223828096 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 18:27:21.007963 123946223828096 attentions.py:1156] attentions/value Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0416 18:27:21.030168 123946223828096 attentions.py:1197] attentions/out Logical: bfloat16[8,2048,16,128]..................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv').
I0416 18:27:21.030231 123946223828096 attentions.py:1197] attentions/out Physical: bfloat16[8,2048,16,128]..................................... ('fsdp', None, None, None).
I0416 18:27:21.074917 123946223828096 linears.py:525] linears/x Logical: bfloat16[8,2048,7168]....................................... ('activation_batch', 'activation_length', 'activation_mlp').
I0416 18:27:21.074981 123946223828096 linears.py:525] linears/x Physical: bfloat16[8,2048,7168]....................................... ('fsdp', None, None).
I0416 18:27:21.457087 123946223828096 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint
I0416 18:27:21.457185 123946223828096 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint.
W0416 18:27:21.524723 3981303 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 18:27:22.289554 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.289704 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.289744 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.289775 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.289801 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.289825 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_0/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.289849 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.289871 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.289892 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.289914 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.289936 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.289958 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wi_1/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.289979 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290001 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290022 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290057 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290081 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290103 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/mlp/wo/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290123 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290143 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290163 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290183 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290202 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290222 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/key/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290241 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290262 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290282 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290301 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290320 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290340 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/out/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290360 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290380 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290400 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290419 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290439 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290458 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/query/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290478 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290498 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/input_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290517 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290536 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/kernel_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290555 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_amax_history
Shape: float32[16,1024]
Logical: P()
Physical: ()
I0416 18:27:22.290574 123946223828096 maxtext_utils.py:1796] params/_overwrite_with_gradient/decoder/layers/self_attention/value/Fp8DirectDotGeneralOp_0/output_grad_scale
Shape: float32[16,1]
Logical: P()
Physical: ()
I0416 18:27:22.290618 123946223828096 maxtext_utils.py:1796] params/params/decoder/decoder_norm/scale
Shape: float32[2048]
Logical: P('norm',)
Physical: (None,)
I0416 18:27:22.290662 123946223828096 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_0/kernel
Shape: float32[2048,16,7168]
Logical: P('embed', 'layers', 'mlp')
Physical: ('fsdp', None, None)
I0416 18:27:22.290689 123946223828096 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wi_1/kernel
Shape: float32[2048,16,7168]
Logical: P('embed', 'layers', 'mlp')
Physical: ('fsdp', None, None)
I0416 18:27:22.290724 123946223828096 maxtext_utils.py:1796] params/params/decoder/layers/mlp/wo/kernel
Shape: float32[7168,16,2048]
Logical: P('mlp', 'layers', 'embed')
Physical: (None, None, 'fsdp')
I0416 18:27:22.290758 123946223828096 maxtext_utils.py:1796] params/params/decoder/layers/post_self_attention_layer_norm/scale
Shape: float32[2048,16]
Logical: P('norm', 'layers')
Physical: (None, None)
I0416 18:27:22.290782 123946223828096 maxtext_utils.py:1796] params/params/decoder/layers/pre_self_attention_layer_norm/scale
Shape: float32[2048,16]
Logical: P('norm', 'layers')
Physical: (None, None)
I0416 18:27:22.290818 123946223828096 maxtext_utils.py:1796] 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 18:27:22.290859 123946223828096 maxtext_utils.py:1796] 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 18:27:22.290886 123946223828096 maxtext_utils.py:1796] 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 18:27:22.290911 123946223828096 maxtext_utils.py:1796] 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 18:27:22.290945 123946223828096 maxtext_utils.py:1796] params/params/decoder/logits_dense/kernel
Shape: float32[2048,32000]
Logical: P('embed_vocab', 'vocab')
Physical: ('fsdp', None)
I0416 18:27:22.290977 123946223828096 maxtext_utils.py:1796] params/params/token_embedder/embedding
Shape: float32[32000,2048]
Logical: P('vocab', 'embed_vocab')
Physical: (None, 'fsdp')
I0416 18:27:23.701390 123946223828096 train.py:157] train/xent Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0416 18:27:23.701472 123946223828096 train.py:157] train/xent Physical: float32[8,2048]............................................. ('fsdp', None).
I0416 18:27:23.716090 123946223828096 train.py:164] train/z_loss Logical: float32[8,2048]............................................. ('activation_embed_and_logits_batch', 'activation_length').
I0416 18:27:23.716143 123946223828096 train.py:164] train/z_loss Physical: float32[8,2048]............................................. ('fsdp', None).
W0416 18:27:24.194662 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 18:27:24.386136 123946223828096 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 18:27:24.386626 123946223828096 max_utils.py:194] tensorboardX not available; using no-op SummaryWriter.
I0416 18:27:24.386827 123946223828096 metric_logger.py:289] number parameters: 1.105 billion
W0416 18:27:28.679068 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 18:27:28.892382 123946223828096 checkpointing.py:794] Waiting for step 0 to finish before checkpoint...
I0416 18:27:29.005598 123946223828096 checkpointing.py:798] Waited 0.1132044792175293 seconds for step 0 to finish before starting checkpointing.
I0416 18:27:29.006088 123946223828096 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0416 18:27:29.006238 123946223828096 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 0
I0416 18:27:29.006599 123946223828096 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0.
I0416 18:27:29.093255 123946223828096 signaling_client.py:373] Using ThreadSafeKeyValueSignalingClient
I0416 18:27:29.181189 123832752932416 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0
I0416 18:27:29.196344 123946223828096 jax_array_handlers.py:347] Scheduling D2H of 81 prioritized jax.Array.
I0416 18:27:29.196437 123946223828096 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
W0416 18:27:29.865901 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 18:27:29.880856 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 18:27:29.889559 123832742446656 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0/items
W0416 18:27:29.901947 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 18:27:29.914938 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 18:27:29.921157 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 18:27:29.925549 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 18:27:29.929442 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 18:27:29.933335 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 18:27:30.072733 123832710989376 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776364049788091724, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776364049788091724, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0/_CHECKPOINT_METADATA
I0416 18:27:30.194457 123946223828096 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.998569s
I0416 18:27:30.194900 123946223828096 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 11.214 GiB/s (total gbytes: 12.3 GiB) (time elapsed: a second) (per-host)
I0416 18:27:30.194950 123946223828096 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 1.100705s (batch_requests_ready=0.094764s, total_serialization_initiated=1.005603s, others=0.000338s)
I0416 18:27:30.195029 123946223828096 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 1.101290s (all_items=0.000019s, per_item={'items': '0.00001884'}, temp_paths=1.101271)
I0416 18:27:30.195825 123832690017856 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0416 18:27:30.195929 123946223828096 async_checkpointer.py:561] Finished blocking save. Time taken: 1.189652s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0.
I0416 18:27:30.196129 123946223828096 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize
I0416 18:27:30.196300 123832763418176 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0416 18:27:30.196409 123946223828096 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776364049.0060744, 'wait_for_prev_duration_secs': 4.2438507080078125e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776364049.0062587, 'checkpointer_blocking_duration_secs': 1.1897945404052734, 'get_old_steps_start_time': 1776364050.196074, 'get_old_steps_duration_secs': 2.2172927856445312e-05, 'checkpoint_manager_blocking_start_time': 1776364049.0059764, 'checkpoint_manager_blocking_duration_secs': 1.1904065608978271}
I0416 18:27:30.196555 123946223828096 checkpointing.py:409] Started an asynchronous checkpoint save for step 0
I0416 18:27:30.196620 123946223828096 max_utils.py:750]
Memstats: After params initialized:
I0416 18:27:30.196770 123946223828096 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_0(process=0,(0,0,0,0))
I0416 18:27:30.196799 123946223828096 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_1(process=0,(1,0,0,0))
I0416 18:27:30.196818 123946223828096 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_2(process=0,(0,1,0,0))
I0416 18:27:30.196837 123946223828096 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_3(process=0,(1,1,0,0))
I0416 18:27:30.196855 123946223828096 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_4(process=0,(0,2,0,0))
I0416 18:27:30.196872 123946223828096 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_5(process=0,(1,2,0,0))
I0416 18:27:30.196890 123946223828096 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_6(process=0,(0,3,0,0))
I0416 18:27:30.196906 123946223828096 max_utils.py:756] Using (GB) 1.61 / 31.25 (5.152000%) on TPU_7(process=0,(1,3,0,0))
W0416 18:27:30.202639 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 18:27:30.208341 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
W0416 18:27:30.213174 3981303 pjrt_executable.cc:642] Assume version compatibility. PjRt-IFRT does not track XLA executable versions.
I0416 18:27:30.525908 123946223828096 metric_logger.py:185] completed step: 0, seconds: 4.505, TFLOP/s/device: 3.016, Tokens/s/device: 454.598, total_weights: 13328, loss: 10.844
I0416 18:27:30.526981 123946223828096 metric_logger.py:269] To see full metrics 'tensorboard --logdir=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/tensorboard/'
I0416 18:27:30.633600 123946223828096 metric_logger.py:185] completed step: 1, seconds: 1.631, TFLOP/s/device: 8.332, Tokens/s/device: 1255.942, total_weights: 12332, loss: 10.871
I0416 18:27:30.681918 3984073 google_auth_provider.cc:149] Using credentials at ~/.config/gcloud/application_default_credentials.json
I0416 18:27:30.681984 3984073 google_auth_provider.cc:156] Using OAuth2 AuthProvider
I0416 18:27:30.744112 123946223828096 metric_logger.py:185] completed step: 2, seconds: 0.011, TFLOP/s/device: 1233.064, Tokens/s/device: 185860.786, total_weights: 15161, loss: 10.871
I0416 18:27:30.854739 123946223828096 metric_logger.py:185] completed step: 3, seconds: 0.109, TFLOP/s/device: 125.054, Tokens/s/device: 18849.517, total_weights: 13327, loss: 10.857
I0416 18:27:31.455876 123832731960896 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0/items/array_metadatas/process_0
I0416 18:27:54.818723 123946223828096 metric_logger.py:185] completed step: 4, seconds: 0.111, TFLOP/s/device: 122.857, Tokens/s/device: 18518.351, total_weights: 11939, loss: 10.861
I0416 18:27:54.830345 123946223828096 metric_logger.py:185] completed step: 5, seconds: 0.111, TFLOP/s/device: 122.808, Tokens/s/device: 18510.986, total_weights: 15502, loss: 10.882
I0416 18:27:54.938577 123946223828096 metric_logger.py:185] completed step: 6, seconds: 23.965, TFLOP/s/device: 0.567, Tokens/s/device: 85.459, total_weights: 13864, loss: 10.886
I0416 18:27:55.049024 123946223828096 metric_logger.py:185] completed step: 7, seconds: 0.008, TFLOP/s/device: 1805.839, Tokens/s/device: 272195.641, total_weights: 12988, loss: 10.829
I0416 18:27:55.159012 123946223828096 metric_logger.py:185] completed step: 8, seconds: 0.109, TFLOP/s/device: 124.750, Tokens/s/device: 18803.654, total_weights: 13820, loss: 10.848
I0416 18:27:55.165791 123946223828096 checkpointing.py:794] Waiting for step 10 to finish before checkpoint...
I0416 18:27:55.379990 123946223828096 checkpointing.py:798] Waited 0.2141706943511963 seconds for step 10 to finish before starting checkpointing.
I0416 18:27:55.380479 123946223828096 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0416 18:27:56.822637 123832700503616 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 26.136571s (commit=25.710503s, array_metadata_write=0.426068s)
I0416 18:27:56.824002 123832690017856 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 455.761 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 27 seconds) (per-host)
I0416 18:27:56.824128 123832690017856 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 26.628153s.
I0416 18:27:57.131081 123832690017856 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776364049788091724, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0/_CHECKPOINT_METADATA
I0416 18:27:57.313624 123832690017856 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 18:27:57.502207 123832710989376 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776364049788091724, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0/_CHECKPOINT_METADATA
I0416 18:27:57.773462 123832690017856 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0416 18:27:57.774383 123832690017856 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.597351s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0/items
I0416 18:27:57.775063 123832690017856 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0/items
I0416 18:27:58.014315 123832690017856 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0
I0416 18:27:58.693900 123832690017856 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0`.
I0416 18:27:58.694547 123832690017856 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 29.688276s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/0
I0416 18:27:58.694629 123832690017856 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 28.498656s.
I0416 18:27:58.694856 123832763418176 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0416 18:27:58.694964 123832763418176 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0416 18:27:58.695021 123832763418176 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts...
I0416 18:27:58.695082 123832763418176 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts.
I0416 18:27:58.695209 123946223828096 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 18:27:58.695344 123946223828096 checkpoint_manager.py:1441] Waiting for previous save to complete took 3.314891 seconds. If this number is high, consider checkpointing less frequently.
I0416 18:27:58.696477 123946223828096 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 10
I0416 18:27:58.696786 123946223828096 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10.
I0416 18:27:58.866643 123832763418176 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10
I0416 18:27:58.875338 123946223828096 jax_array_handlers.py:347] Scheduling D2H of 81 prioritized jax.Array.
I0416 18:27:58.875436 123946223828096 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 18:27:59.546953 123832547411520 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10/items
I0416 18:28:06.608515 123946223828096 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 7.733619s
I0416 18:28:06.615345 123946223828096 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 1.575 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 7 seconds) (per-host)
I0416 18:28:06.615414 123946223828096 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 7.837720s (batch_requests_ready=0.091156s, total_serialization_initiated=7.739836s, others=0.006728s)
I0416 18:28:06.615498 123946223828096 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 7.838274s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=7.838257)
I0416 18:28:06.616916 123832700503616 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0416 18:28:06.617000 123946223828096 async_checkpointer.py:561] Finished blocking save. Time taken: 7.920481s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10.
I0416 18:28:06.617209 123946223828096 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=10] Starting CheckpointManager Save Finalize thread=save_finalize
I0416 18:28:06.617365 123832690017856 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0416 18:28:06.617475 123946223828096 standard_logger.py:34] {'step': 10, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776364075.3804164, 'wait_for_prev_duration_secs': 3.3148908615112305, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776364078.6964998, 'checkpointer_blocking_duration_secs': 7.920620679855347, 'get_old_steps_start_time': 1776364086.6171358, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1776364075.3803422, 'checkpoint_manager_blocking_duration_secs': 11.237105131149292}
I0416 18:28:06.617611 123946223828096 checkpointing.py:409] Started an asynchronous checkpoint save for step 10
I0416 18:28:06.618040 123946223828096 metric_logger.py:185] completed step: 9, seconds: 0.111, TFLOP/s/device: 122.926, Tokens/s/device: 18528.739, total_weights: 12300, loss: 10.857
I0416 18:28:06.626923 123946223828096 metric_logger.py:185] completed step: 10, seconds: 0.110, TFLOP/s/device: 123.642, Tokens/s/device: 18636.649, total_weights: 13436, loss: 10.876
I0416 18:28:06.734518 123946223828096 metric_logger.py:185] completed step: 11, seconds: 11.459, TFLOP/s/device: 1.186, Tokens/s/device: 178.728, total_weights: 13817, loss: 10.853
I0416 18:28:06.844793 123946223828096 metric_logger.py:185] completed step: 12, seconds: 0.008, TFLOP/s/device: 1693.733, Tokens/s/device: 255297.931, total_weights: 12076, loss: 10.852
I0416 18:28:06.955267 123946223828096 metric_logger.py:185] completed step: 13, seconds: 0.109, TFLOP/s/device: 124.892, Tokens/s/device: 18825.087, total_weights: 11048, loss: 10.864
I0416 18:28:07.065515 123946223828096 metric_logger.py:185] completed step: 14, seconds: 0.110, TFLOP/s/device: 123.216, Tokens/s/device: 18572.426, total_weights: 14267, loss: 10.834
I0416 18:28:07.175822 123946223828096 metric_logger.py:185] completed step: 15, seconds: 0.110, TFLOP/s/device: 123.007, Tokens/s/device: 18540.984, total_weights: 13004, loss: 10.832
I0416 18:28:07.286108 123946223828096 metric_logger.py:185] completed step: 16, seconds: 0.110, TFLOP/s/device: 123.390, Tokens/s/device: 18598.738, total_weights: 13811, loss: 10.867
I0416 18:28:07.374693 123832731960896 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10/items/array_metadatas/process_0
I0416 18:28:07.396387 123946223828096 metric_logger.py:185] completed step: 17, seconds: 0.111, TFLOP/s/device: 122.959, Tokens/s/device: 18533.769, total_weights: 13709, loss: 10.850
I0416 18:28:07.506839 123946223828096 metric_logger.py:185] completed step: 18, seconds: 0.110, TFLOP/s/device: 123.211, Tokens/s/device: 18571.752, total_weights: 12831, loss: 10.839
I0416 18:28:07.617005 123946223828096 checkpointing.py:794] Waiting for step 19 to finish before checkpoint...
I0416 18:28:07.618778 123946223828096 checkpointing.py:798] Waited 0.0017790794372558594 seconds for step 19 to finish before starting checkpointing.
I0416 18:28:07.620738 123946223828096 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=10][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0416 18:28:27.016307 123832763418176 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 20.400232s (commit=20.003991s, array_metadata_write=0.396241s)
I0416 18:28:27.017288 123832700503616 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 447.531 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 28 seconds) (per-host)
I0416 18:28:27.017349 123832700503616 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 20.400301s.
I0416 18:28:27.414708 123832700503616 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 18:28:27.814411 123832700503616 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0416 18:28:27.815176 123832700503616 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.548578s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10/items
I0416 18:28:27.815796 123832700503616 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10/items
I0416 18:28:28.060391 123832700503616 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10
I0416 18:28:28.762616 123832700503616 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10`.
I0416 18:28:28.763242 123832700503616 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 30.066729s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/10
I0416 18:28:28.763310 123832700503616 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 22.146266s.
I0416 18:28:28.763462 123832690017856 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0416 18:28:28.763566 123832690017856 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0416 18:28:28.763620 123832690017856 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is syncing with other hosts...
I0416 18:28:28.763658 123832690017856 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=10] CheckpointManager Save Finalize is done on all hosts.
I0416 18:28:28.763945 123946223828096 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 18:28:28.764073 123946223828096 checkpoint_manager.py:1441] Waiting for previous save to complete took 21.143353 seconds. If this number is high, consider checkpointing less frequently.
I0416 18:28:28.764280 123946223828096 checkpoint_manager.py:1501] [process=0] Saving checkpoint at step 19
I0416 18:28:28.764560 123946223828096 async_checkpointer.py:452] [process=0] Started async saving checkpoint to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19.
I0416 18:28:28.941078 123946223828096 jax_array_handlers.py:347] Scheduling D2H of 81 prioritized jax.Array.
I0416 18:28:28.941166 123946223828096 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 18:28:28.943878 123832690017856 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19
I0416 18:28:30.678306 123832411096640 atomicity.py:137] Creating tmp directory gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19/items
I0416 18:28:32.317480 123946223828096 base_pytree_checkpoint_handler.py:153] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 3.376833s
I0416 18:28:32.324518 123946223828096 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.551 GiB/s (total gbytes: 12.3 GiB) (time elapsed: 3 seconds) (per-host)
I0416 18:28:32.324577 123946223828096 base_pytree_checkpoint_handler.py:732] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 3.475673s (batch_requests_ready=0.085969s, total_serialization_initiated=3.382796s, others=0.006908s)
I0416 18:28:32.324655 123946223828096 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 3.476426s (all_items=0.000012s, per_item={'items': '0.00001240'}, temp_paths=3.476414)
I0416 18:28:32.326133 123832690017856 async_checkpointer.py:79] [process=0][thread=async_save] Background save thread started.
I0416 18:28:32.326217 123946223828096 async_checkpointer.py:561] Finished blocking save. Time taken: 3.561897s. Continuing background save to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19.
I0416 18:28:32.326380 123946223828096 checkpoint_manager.py:1549] [process=0][thread=MainThread][step=19] Starting CheckpointManager Save Finalize thread=save_finalize
I0416 18:28:32.326550 123832669046336 async_checkpointer.py:265] [process=0][thread=save_finalize] Waiting for background save thread=async_save.
I0416 18:28:32.326636 123946223828096 standard_logger.py:34] {'step': 19, 'event_type': 'save', 'directory': 'gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776364087.6207044, 'wait_for_prev_duration_secs': 21.14335322380066, 'time_between_consecutive_saves_sec': 8.925606727600098, 'checkpointer_blocking_start_time': 1776364108.7643023, 'checkpointer_blocking_duration_secs': 3.562019109725952, 'get_old_steps_start_time': 1776364112.3263316, 'get_old_steps_duration_secs': 1.7642974853515625e-05, 'checkpoint_manager_blocking_start_time': 1776364087.6206615, 'checkpoint_manager_blocking_duration_secs': 24.705949544906616}
I0416 18:28:32.326723 123946223828096 checkpointing.py:409] Started an asynchronous checkpoint save for step 19
I0416 18:28:32.326754 123946223828096 checkpoint_manager.py:1994] [process=0][thread=MainThread][step=19][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete.
I0416 18:28:33.126320 123832763418176 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 81 array_metadata.ArrayMetadata to gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19/items/array_metadatas/process_0
I0416 18:29:04.175814 123832411096640 base_pytree_checkpoint_handler.py:1217] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 31.850546s (commit=31.391596s, array_metadata_write=0.458950s)
I0416 18:29:04.176806 123832690017856 base_pytree_checkpoint_handler.py:128] [process=0] /jax/checkpoint/write/gbytes_per_sec: 357.737 MiB/s (total gbytes: 12.3 GiB) (time elapsed: 35 seconds) (per-host)
I0416 18:29:04.176857 123832690017856 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 31.850596s.
I0416 18:29:04.597850 123832690017856 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 18:29:05.031183 123832690017856 ocdbt_utils.py:56] Param validation support for Zarr3 will be added later (b/362328389).
I0416 18:29:05.031910 123832690017856 base_pytree_checkpoint_handler.py:1342] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 0.583111s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19/items
I0416 18:29:05.032576 123832690017856 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19/items
I0416 18:29:05.276296 123832690017856 atomicity.py:608] Finalizing gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19
I0416 18:29:05.933298 123832690017856 atomicity.py:794] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19`.
I0416 18:29:05.933921 123832690017856 async_checkpointer.py:420] Finished async_save (blocking + background). Time taken: 37.169607s. directory=gs://wanglance-maxtext/linen_ckpt_feat_nnx_post_train_fixes_20260416_181521/linen_feat_nnx_post_train_fixes_20260416_181521_05_fp8/checkpoints/19
I0416 18:29:05.933982 123832690017856 async_checkpointer.py:144] [process=0][thread=async_save] Background save thread done. Time taken: 33.607721s.
I0416 18:29:05.934106 123832669046336 async_checkpointer.py:273] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save.
I0416 18:29:05.934155 123832669046336 async_checkpointer.py:283] [process=0][thread=save_finalize] No errors found in background save thread=async_save.
I0416 18:29:05.934196 123832669046336 checkpoint_manager.py:2103] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is syncing with other hosts...
I0416 18:29:05.934231 123832669046336 checkpoint_manager.py:2112] [process=0][thread=save_finalize][step=19] CheckpointManager Save Finalize is done on all hosts.
I0416 18:29:05.934761 123946223828096 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 18:29:05.935358 123946223828096 checkpoint_manager.py:1983] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0416 18:29:05.936120 123946223828096 metric_logger.py:185] completed step: 19, seconds: 0.110, TFLOP/s/device: 123.390, Tokens/s/device: 18598.738, 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