main (574ad3fb9) vs feat/nnx-post-train-fixes (574ad3fb9)| Metric | main 574ad3fb9 | feat/nnx-post-train-fixes 574ad3fb9 | Diff (feat/nnx-post-train-fixes − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 8.1790 | 8.1790 | 0 |
| TFLOP/s | 92.288 | 92.250 | -0.038 |
| Tok/s | 13910.6 | 13904.9 | -5.666 |
| Avg s/step | 2.893 | 2.766 | -0.127 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Sat Apr 18 19:07:40 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-18 19:08:04.016859: E external/local_xla/xla/stream_executor/cuda/cuda_platform.cc:51] failed call to cuInit: INTERNAL: CUDA error: Failed call to cuInit: UNKNOWN ERROR (303) I0418 19:08:04.544170 134895158245184 max_utils.py:273] Attempting to initialize the jax distributed system... I0418 19:08:13.585237 134895158245184 distributed.py:149] Starting JAX distributed service on [::]:8482 I0418 19:08:13.587556 134895158245184 distributed.py:172] Connecting to JAX distributed service on mt-13-scan-layers-true-fmgzw-slice-job-0-0.mt-13-scan-layers-true-fmgzw:8482 I0418 19:08:14.777020 134895158245184 max_utils.py:284] Jax distributed system initialized! I0418 19:08:20.959322 134895158245184 max_utils.py:800] System Information: Jax Version: 0.9.2 I0418 19:08:20.959425 134895158245184 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0418 19:08:20.959484 134895158245184 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0418 19:08:20.959522 134895158245184 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0418 19:08:21.648977 134895158245184 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0418 19:08:21.649245 134895158245184 checkpointing.py:677] Setting up checkpoint logger... I0418 19:08:21.649300 134895158245184 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0418 19:08:21.649343 134895158245184 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0418 19:08:21.649681 134895158245184 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7aaf20f141a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0418 19:08:24.951020 134895158245184 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0418 19:08:24.951252 134895158245184 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a9acc487bf0>}, handler_registry=None I0418 19:08:24.951501 134895158245184 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a9acc487bf0>` for item "items" and save args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>` to `_handler_registry`. I0418 19:08:24.951549 134895158245184 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a9acc32f5c0>` for item "metrics" and save args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>` and restore args `<class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>` to `_handler_registry`. I0418 19:08:24.951584 134895158245184 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 0x7a9acc487bf0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a9acc487bf0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a9acc32f5c0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a9acc32f5c0>}). I0418 19:08:24.951901 134895158245184 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0418 19:08:24.951973 134895158245184 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7a99c45ff880> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0418 19:08:26.987023 134895158245184 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints I0418 19:08:27.855195 134895158245184 checkpoint_manager.py:929] [process=5][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a9acc32bc50> I0418 19:08:27.855361 134895158245184 checkpointing.py:301] Checkpoint manager created! I0418 19:08:28.780652 134895158245184 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0418 19:08:28.780764 134895158245184 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:08:29.157340 134895158245184 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:08:29.157423 134895158245184 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:08:29.173671 134895158245184 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0418 19:08:29.173735 134895158245184 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0418 19:08:29.197433 134895158245184 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 19:08:29.197514 134895158245184 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:08:29.213881 134895158245184 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 19:08:29.213942 134895158245184 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:08:29.230174 134895158245184 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0418 19:08:29.230232 134895158245184 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:08:29.255619 134895158245184 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0418 19:08:29.255688 134895158245184 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0418 19:08:29.276583 134895158245184 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0418 19:08:29.276648 134895158245184 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0418 19:08:29.493539 134895158245184 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0418 19:08:29.493644 134895158245184 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0418 19:08:30.907189 134895158245184 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0418 19:08:30.907311 134895158245184 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 19:08:30.907362 134895158245184 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0418 19:08:30.907418 134895158245184 maxtext_utils.py:1654] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0418 19:08:30.907486 134895158245184 maxtext_utils.py:1654] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 19:08:30.907528 134895158245184 maxtext_utils.py:1654] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0418 19:08:30.907579 134895158245184 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0418 19:08:30.907629 134895158245184 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: P('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0418 19:08:30.907678 134895158245184 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0418 19:08:30.907720 134895158245184 maxtext_utils.py:1654] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: P('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0418 19:08:30.907767 134895158245184 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0418 19:08:30.907814 134895158245184 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0418 19:08:31.397019 134895158245184 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:08:31.397113 134895158245184 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:08:31.412441 134895158245184 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0418 19:08:31.412510 134895158245184 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0418 19:08:42.264530 134895158245184 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0418 19:08:42.265280 134895158245184 metric_logger.py:301] number parameters: 1.104 billion I0418 19:08:53.581965 134895158245184 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0418 19:08:53.735486 134895158245184 checkpointing.py:776] Waited 0.15348434448242188 seconds for step 0 to finish before starting checkpointing. I0418 19:08:53.737769 134895158245184 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:08:53.739607 134895158245184 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0 I0418 19:08:53.741130 134895158245184 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints/0. I0418 19:08:54.492208 134895158245184 signaling_client.py:364] Using JaxDistributedSignalingClient I0418 19:08:54.493223 134895158245184 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:08:54.493281 134895158245184 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0418 19:08:54.767407 134895158245184 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.275273s I0418 19:08:54.767588 134895158245184 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.496 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28068017959594727 s) (per-host) I0418 19:08:54.767643 134895158245184 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.280746s (batch_requests_ready=0.002210s, total_serialization_initiated=0.278462s, others=0.000073s) I0418 19:08:54.767745 134895158245184 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.284901s (all_items=0.000018s, per_item={'items': '0.00001764'}, temp_paths=0.284883) I0418 19:08:54.768565 134895158245184 event_tracking.py:125] [process=5] [async] Finished blocking save in 1.03 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints/0. I0418 19:08:54.768901 134766141634304 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 19:28:54.768862 I0418 19:08:54.780236 134895158245184 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:08:54.780526 134765025113856 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:08:54.780660 134895158245184 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776539333.7377515, 'wait_for_prev_duration_secs': 5.984306335449219e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776539333.739645, 'checkpointer_blocking_duration_secs': 1.0294127464294434, 'get_old_steps_start_time': 1776539334.7690845, 'get_old_steps_duration_secs': 2.9087066650390625e-05, 'checkpoint_manager_blocking_start_time': 1776539333.7358968, 'checkpoint_manager_blocking_duration_secs': 1.0447235107421875} I0418 19:08:54.780775 134895158245184 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0418 19:08:54.780824 134895158245184 max_utils.py:750] Memstats: After params initialized: I0418 19:08:54.780874 134895158245184 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0418 19:08:54.780907 134895158245184 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0418 19:08:54.780931 134895158245184 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0418 19:08:54.780955 134895158245184 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0418 19:08:55.092705 134895158245184 metric_logger.py:196] completed step: 0, seconds: 11.317, TFLOP/s/device: 1.201, Tokens/s/device: 180.973, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0418 19:08:55.288127 134895158245184 metric_logger.py:196] completed step: 1, seconds: 1.509, TFLOP/s/device: 9.003, Tokens/s/device: 1357.011, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0418 19:08:55.696110 134895158245184 metric_logger.py:196] completed step: 2, seconds: 0.048, TFLOP/s/device: 281.336, Tokens/s/device: 42406.046, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0418 19:08:55.843300 134895158245184 metric_logger.py:196] completed step: 3, seconds: 0.408, TFLOP/s/device: 33.284, Tokens/s/device: 5016.890, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0418 19:08:56.138497 134895158245184 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.929, Tokens/s/device: 13404.370, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0418 19:08:56.144357 134895158245184 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.328, Tokens/s/device: 13916.636, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0418 19:08:59.554563 2874 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0418 19:09:01.896274 134765067892480 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints/0/items/array_metadatas/process_5 I0418 19:09:19.469968 134895158245184 metric_logger.py:196] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.948, Tokens/s/device: 6925.774, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0418 19:09:19.617210 134895158245184 metric_logger.py:196] completed step: 7, seconds: 23.179, TFLOP/s/device: 0.586, Tokens/s/device: 88.356, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0418 19:09:19.765043 134895158245184 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.284, Tokens/s/device: 13457.924, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0418 19:09:19.911949 134895158245184 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0418 19:09:19.912597 134895158245184 checkpointing.py:776] Waited 0.0006701946258544922 seconds for step 9 to finish before starting checkpointing. I0418 19:09:19.914944 134895158245184 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:09:30.793346 134766141634304 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 43.507 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 36.306397438049316 s) (per-host) I0418 19:09:30.793490 134766141634304 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 36.024473s. I0418 19:09:41.430649 134766141634304 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 46.661617s. I0418 19:09:41.430930 134765025113856 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:09:41.431061 134765025113856 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:09:41.431111 134765025113856 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:09:41.432601 134765025113856 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0418 19:09:41.432762 134895158245184 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0418 19:09:41.432891 134895158245184 checkpoint_manager.py:1452] Waiting for previous save to complete took 21.517948 seconds. If this number is high, consider checkpointing less frequently. I0418 19:09:41.434678 134895158245184 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9 I0418 19:09:41.436726 134895158245184 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints/9. I0418 19:09:42.163845 134895158245184 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0418 19:09:42.163940 134895158245184 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0418 19:09:42.199553 134895158245184 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036779s I0418 19:09:42.199711 134895158245184 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 38.419 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.040151119232177734 s) (per-host) I0418 19:09:42.199758 134895158245184 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040207s (batch_requests_ready=0.001735s, total_serialization_initiated=0.038409s, others=0.000062s) I0418 19:09:42.199840 134895158245184 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044278s (all_items=0.000033s, per_item={'items': '0.00003266'}, temp_paths=0.044245) I0418 19:09:42.200522 134895158245184 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.77 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints/9. I0418 19:09:42.200856 134765025113856 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-18 19:29:42.200819 I0418 19:09:42.202786 134895158245184 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0418 19:09:42.203045 134741260998400 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0418 19:09:42.203207 134895158245184 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776539359.9149146, 'wait_for_prev_duration_secs': 21.517947912216187, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776539381.434716, 'checkpointer_blocking_duration_secs': 0.7662858963012695, 'get_old_steps_start_time': 1776539382.2010262, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776539359.9128253, 'checkpoint_manager_blocking_duration_secs': 22.290345430374146} I0418 19:09:42.203318 134895158245184 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0418 19:09:42.203365 134895158245184 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0418 19:09:50.357089 134765067892480 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260418_180002/linen_xpk_main_20260418_180002_13_scan_layers_true/checkpoints/9/items/array_metadatas/process_5 I0418 19:10:25.263633 134765025113856 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 36.646 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 43.104031801223755 s) (per-host) I0418 19:10:25.263773 134765025113856 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 43.062795s. I0418 19:10:35.901714 134765025113856 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 53.700731s. I0418 19:10:35.902004 134741260998400 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0418 19:10:35.902141 134741260998400 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0418 19:10:35.902203 134741260998400 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0418 19:10:35.903820 134741260998400 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0418 19:10:35.903981 134895158245184 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0418 19:10:35.904128 134895158245184 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0418 19:10:35.905076 134895158245184 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.288, Tokens/s/device: 13910.586, total_weights: 65536, loss: 8.179, lm_loss: 8.179, perplexity: 3564.635 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sat Apr 18 19:10:46 UTC 2026 EXIT_CODE=0
XPK Start: Sun Apr 19 00:31:36 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-19 00:32:00.552040: 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) I0419 00:32:00.730060 140516397336384 max_utils.py:273] Attempting to initialize the jax distributed system... I0419 00:32:09.771401 140516397336384 distributed.py:149] Starting JAX distributed service on [::]:8482 I0419 00:32:09.773622 140516397336384 distributed.py:172] Connecting to JAX distributed service on mt-13-scan-layers-true-xr4ns-slice-job-0-0.mt-13-scan-layers-true-xr4ns:8482 I0419 00:32:11.370967 140516397336384 max_utils.py:284] Jax distributed system initialized! I0419 00:32:17.373416 140516397336384 max_utils.py:800] System Information: Jax Version: 0.9.2 I0419 00:32:17.373521 140516397336384 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0419 00:32:17.373561 140516397336384 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Mar 4 2026 11:32:08 (1772652728) cl/878335365 I0419 00:32:17.373600 140516397336384 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0419 00:32:18.071426 140516397336384 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 00:32:18.072012 140516397336384 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0419 00:32:18.072193 140516397336384 checkpointing.py:688] Setting up checkpoint logger... I0419 00:32:18.072257 140516397336384 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0419 00:32:18.072302 140516397336384 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0419 00:32:18.072642 140516397336384 base_pytree_checkpoint_handler.py:441] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=True, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7fcbe4f18410>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0419 00:32:23.192623 140516397336384 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0419 00:32:23.192855 140516397336384 checkpoint_manager.py:708] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fb7bc19b860>}, handler_registry=None I0419 00:32:23.193092 140516397336384 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fb7bc19b860>` 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`. I0419 00:32:23.193139 140516397336384 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fb6e479d6a0>` 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`. I0419 00:32:23.193176 140516397336384 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 0x7fb7bc19b860>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7fb7bc19b860>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fb6e479d6a0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7fb6e479d6a0>}). I0419 00:32:23.193515 140516397336384 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0419 00:32:23.193584 140516397336384 async_checkpointer.py:192] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7fb6e43d7740> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0419 00:32:23.867935 140516397336384 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints I0419 00:32:24.815894 140516397336384 checkpoint_manager.py:929] [process=6][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7fb6e4798e90> I0419 00:32:24.816070 140516397336384 checkpointing.py:302] Checkpoint manager created! I0419 00:32:25.748690 140516397336384 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0419 00:32:25.748797 140516397336384 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 00:32:26.129353 140516397336384 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 00:32:26.129449 140516397336384 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 00:32:26.145767 140516397336384 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0419 00:32:26.145836 140516397336384 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0419 00:32:26.169412 140516397336384 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 00:32:26.169487 140516397336384 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 00:32:26.185915 140516397336384 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 00:32:26.185988 140516397336384 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 00:32:26.202180 140516397336384 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0419 00:32:26.202262 140516397336384 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 00:32:26.226816 140516397336384 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0419 00:32:26.226892 140516397336384 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0419 00:32:26.247707 140516397336384 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0419 00:32:26.247780 140516397336384 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0419 00:32:26.463079 140516397336384 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0419 00:32:26.463188 140516397336384 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0419 00:32:27.905970 140516397336384 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0419 00:32:27.906094 140516397336384 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 00:32:27.906147 140516397336384 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0419 00:32:27.906203 140516397336384 maxtext_utils.py:1827] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0419 00:32:27.906270 140516397336384 maxtext_utils.py:1827] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0419 00:32:27.906311 140516397336384 maxtext_utils.py:1827] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0419 00:32:27.906363 140516397336384 maxtext_utils.py:1827] 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) I0419 00:32:27.906416 140516397336384 maxtext_utils.py:1827] 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') I0419 00:32:27.906455 140516397336384 maxtext_utils.py:1827] 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) I0419 00:32:27.906491 140516397336384 maxtext_utils.py:1827] 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) I0419 00:32:27.906543 140516397336384 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0419 00:32:27.906593 140516397336384 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0419 00:32:28.397409 140516397336384 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 00:32:28.397501 140516397336384 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0419 00:32:28.412830 140516397336384 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0419 00:32:28.412893 140516397336384 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0419 00:32:39.491880 140516397336384 max_utils.py:791] Total memory size: 1.5 GB, Output size: 0.4 GB, Temp size: 1.1 GB, Argument size: 0.4 GB, Host temp size: 0.0 GB. I0419 00:32:39.492611 140516397336384 metric_logger.py:301] number parameters: 1.104 billion I0419 00:32:50.776613 140516397336384 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0419 00:32:50.930047 140516397336384 checkpointing.py:798] Waited 0.15341591835021973 seconds for step 0 to finish before starting checkpointing. I0419 00:32:50.932353 140516397336384 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 00:32:50.934028 140516397336384 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 0 I0419 00:32:50.935451 140516397336384 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints/0. I0419 00:32:51.666898 140516397336384 signaling_client.py:364] Using JaxDistributedSignalingClient I0419 00:32:51.667816 140516397336384 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 00:32:51.667875 140516397336384 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 00:32:51.945367 140516397336384 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.278545s I0419 00:32:51.945545 140516397336384 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 5.433 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28393983840942383 s) (per-host) I0419 00:32:51.945602 140516397336384 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.284006s (batch_requests_ready=0.002181s, total_serialization_initiated=0.281750s, others=0.000075s) I0419 00:32:51.945701 140516397336384 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.288085s (all_items=0.000018s, per_item={'items': '0.00001812'}, temp_paths=0.288067) I0419 00:32:51.946526 140516397336384 event_tracking.py:125] [process=6] [async] Finished blocking save in 1.01 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints/0. I0419 00:32:51.946863 140388226324224 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 00:52:51.946826 I0419 00:32:51.954807 140516397336384 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0419 00:32:51.955080 140386076976896 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0419 00:32:51.955247 140516397336384 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776558770.9323332, 'wait_for_prev_duration_secs': 6.008148193359375e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776558770.9340672, 'checkpointer_blocking_duration_secs': 1.0129468441009521, 'get_old_steps_start_time': 1776558771.9470413, 'get_old_steps_duration_secs': 3.0517578125e-05, 'checkpoint_manager_blocking_start_time': 1776558770.93056, 'checkpoint_manager_blocking_duration_secs': 1.0246365070343018} I0419 00:32:51.955359 140516397336384 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0419 00:32:51.955411 140516397336384 max_utils.py:750] Memstats: After params initialized: I0419 00:32:51.955461 140516397336384 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_24(process=6,(0,6,0,0)) I0419 00:32:51.955494 140516397336384 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_25(process=6,(1,6,0,0)) I0419 00:32:51.955523 140516397336384 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_28(process=6,(0,7,0,0)) I0419 00:32:51.955549 140516397336384 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_29(process=6,(1,7,0,0)) I0419 00:32:52.270311 140516397336384 metric_logger.py:196] completed step: 0, seconds: 11.284, TFLOP/s/device: 1.204, Tokens/s/device: 181.498, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 00:32:52.444163 140516397336384 metric_logger.py:196] completed step: 1, seconds: 1.492, TFLOP/s/device: 9.106, Tokens/s/device: 1372.530, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0419 00:32:52.843768 140516397336384 metric_logger.py:196] completed step: 2, seconds: 0.027, TFLOP/s/device: 508.500, Tokens/s/device: 76646.707, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0419 00:32:52.991138 140516397336384 metric_logger.py:196] completed step: 3, seconds: 0.400, TFLOP/s/device: 33.979, Tokens/s/device: 5121.652, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0419 00:32:53.287596 140516397336384 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.848, Tokens/s/device: 13392.098, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0419 00:32:53.293563 140516397336384 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.273, Tokens/s/device: 13908.414, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0419 00:32:56.113187 2855 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0419 00:32:59.900738 140386589161216 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints/0/items/array_metadatas/process_6 I0419 00:33:15.519978 140516397336384 metric_logger.py:196] completed step: 6, seconds: 0.297, TFLOP/s/device: 45.742, Tokens/s/device: 6894.764, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0419 00:33:15.667191 140516397336384 metric_logger.py:196] completed step: 7, seconds: 22.080, TFLOP/s/device: 0.615, Tokens/s/device: 92.755, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0419 00:33:15.814490 140516397336384 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.329, Tokens/s/device: 13464.649, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0419 00:33:15.961328 140516397336384 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0419 00:33:15.961954 140516397336384 checkpointing.py:798] Waited 0.0006556510925292969 seconds for step 9 to finish before starting checkpointing. I0419 00:33:15.964012 140516397336384 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 00:33:29.084940 140388226324224 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 42.209 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.42328143119812 s) (per-host) I0419 00:33:29.085063 140388226324224 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.138086s. I0419 00:33:39.044190 140388226324224 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 47.097196s. I0419 00:33:39.044419 140386076976896 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 00:33:39.044476 140386076976896 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0419 00:33:39.044533 140386076976896 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0419 00:33:39.046529 140386076976896 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0419 00:33:39.046687 140516397336384 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0419 00:33:39.046825 140516397336384 checkpoint_manager.py:1452] Waiting for previous save to complete took 23.082813 seconds. If this number is high, consider checkpointing less frequently. I0419 00:33:39.048637 140516397336384 checkpoint_manager.py:1512] [process=6] Saving checkpoint at step 9 I0419 00:33:39.050704 140516397336384 event_tracking.py:70] [process=6] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints/9. I0419 00:33:39.774292 140516397336384 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0419 00:33:39.774388 140516397336384 replica_slices.py:424] Transferring arrays to host memory with options: use_replica_parallel=True, min_slice_bytes_for_replica_parallel=None, max_replicas_for_replica_parallel=None, enable_pinned_host_transfer=False I0419 00:33:39.809089 140516397336384 base_pytree_checkpoint_handler.py:154] [process=6][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036013s I0419 00:33:39.809257 140516397336384 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/blocking_gbytes_per_sec: 39.080 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.039472341537475586 s) (per-host) I0419 00:33:39.809310 140516397336384 base_pytree_checkpoint_handler.py:768] [process=6][thread=MainThread] Initiated Pytree async_save. Time taken: 0.039546s (batch_requests_ready=0.001816s, total_serialization_initiated=0.037649s, others=0.000081s) I0419 00:33:39.809405 140516397336384 composite_checkpoint_handler.py:715] [process=6][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043755s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.043738) I0419 00:33:39.810059 140516397336384 event_tracking.py:125] [process=6] [async] Finished blocking save in 0.76 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints/9. I0419 00:33:39.810401 140386076976896 async_checkpointer.py:76] [process=6][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-19 00:53:39.810364 I0419 00:33:39.812366 140516397336384 checkpoint_manager.py:1560] [process=6][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0419 00:33:39.812601 140386589161216 async_checkpointer.py:280] [process=6][thread=save_finalize] Waiting for background save thread=async_save. I0419 00:33:39.812732 140516397336384 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776558795.9639816, 'wait_for_prev_duration_secs': 23.082812547683716, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776558819.0486774, 'checkpointer_blocking_duration_secs': 0.7618741989135742, 'get_old_steps_start_time': 1776558819.8105772, 'get_old_steps_duration_secs': 3.075599670410156e-05, 'checkpoint_manager_blocking_start_time': 1776558795.9621737, 'checkpoint_manager_blocking_duration_secs': 23.850522756576538} I0419 00:33:39.812843 140516397336384 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0419 00:33:39.812887 140516397336384 checkpoint_manager.py:2020] [process=6][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0419 00:33:46.079335 140365948503808 array_metadata_store.py:203] [process=6][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260418_235007/linen_xpk_feat_nnx_post_train_fixes_20260418_235007_13_scan_layers_true/checkpoints/9/items/array_metadatas/process_6 I0419 00:34:20.564414 140386076976896 base_pytree_checkpoint_handler.py:130] [process=6] /jax/orbax/write/gbytes_per_sec: 38.721 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.79461336135864 s) (per-host) I0419 00:34:20.564521 140386076976896 async_checkpointer.py:90] [process=6][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.754005s. I0419 00:34:30.617685 140386076976896 async_checkpointer.py:160] [process=6][thread=async_save] Background save thread done. Time taken: 50.807154s. I0419 00:34:30.617963 140386589161216 async_checkpointer.py:288] [process=6][thread=save_finalize] Done with waiting for background save thread=async_save. I0419 00:34:30.618087 140386589161216 async_checkpointer.py:298] [process=6][thread=save_finalize] No errors found in background save thread=async_save. I0419 00:34:30.618137 140386589161216 checkpoint_manager.py:2137] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0419 00:34:30.619900 140386589161216 checkpoint_manager.py:2146] [process=6][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0419 00:34:30.620076 140516397336384 checkpoint_manager.py:2032] [process=6][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0419 00:34:30.620224 140516397336384 checkpoint_manager.py:2009] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0419 00:34:30.621255 140516397336384 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.250, Tokens/s/device: 13904.920, total_weights: 65536, loss: 8.179, lm_loss: 8.179, perplexity: 3564.635 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Sun Apr 19 00:34:41 UTC 2026 EXIT_CODE=0