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 | 83.427 | 83.425 | -0.002 |
| Tok/s | 12575.0 | 12574.7 | -0.309 |
| Avg s/step | 2.661 | 2.783 | +0.122 |
| Memory % | 0.54 | 0.54 | 0 |
| JAX | 0.9.2 | 0.9.2 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Mon Apr 20 06:40:36 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-20 06:41:00.279447: 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) I0420 06:41:00.816740 134829274994496 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 06:41:09.857061 134829274994496 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 06:41:09.859384 134829274994496 distributed.py:172] Connecting to JAX distributed service on mt-11-optimizer-offload-nzll1-slice-job-0-0.mt-11-optimizer-offload-nzll1:8482 I0420 06:41:11.025671 134829274994496 max_utils.py:284] Jax distributed system initialized! I0420 06:41:16.150786 134829274994496 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 06:41:16.150887 134829274994496 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 06:41:16.150928 134829274994496 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 I0420 06:41:16.150964 134829274994496 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 06:41:16.843600 134829274994496 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 06:41:16.843880 134829274994496 checkpointing.py:677] Setting up checkpoint logger... I0420 06:41:16.843935 134829274994496 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 06:41:16.843981 134829274994496 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 06:41:16.844316 134829274994496 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 0x7a9fc011f9e0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 06:41:19.817302 134829274994496 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0420 06:41:19.817549 134829274994496 checkpoint_manager.py:708] [process=4][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a8c40655f40>}, handler_registry=None I0420 06:41:19.817788 134829274994496 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a8c40655f40>` 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`. I0420 06:41:19.817839 134829274994496 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a8c40658fb0>` 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`. I0420 06:41:19.817874 134829274994496 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 0x7a8c40655f40>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7a8c40655f40>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a8c40658fb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7a8c40658fb0>}). I0420 06:41:19.818189 134829274994496 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 06:41:19.818257 134829274994496 async_checkpointer.py:192] [process=4][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7a8c2825b4c0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 06:41:20.805624 134829274994496 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints I0420 06:41:20.954488 134829274994496 checkpoint_manager.py:929] [process=4][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_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7a8c406555e0> I0420 06:41:20.954665 134829274994496 checkpointing.py:301] Checkpoint manager created! I0420 06:41:21.874707 134829274994496 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 06:41:21.874825 134829274994496 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:41:22.257097 134829274994496 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:41:22.257191 134829274994496 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:41:22.273482 134829274994496 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:41:22.273547 134829274994496 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:41:22.297097 134829274994496 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 06:41:22.297160 134829274994496 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:41:22.313546 134829274994496 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 06:41:22.313605 134829274994496 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:41:22.329729 134829274994496 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 06:41:22.329787 134829274994496 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:41:22.355256 134829274994496 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 06:41:22.355346 134829274994496 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:41:22.376439 134829274994496 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 06:41:22.376502 134829274994496 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 06:41:22.596749 134829274994496 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0420 06:41:22.596859 134829274994496 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 06:41:24.196776 134829274994496 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 06:41:24.196907 134829274994496 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) I0420 06:41:24.196964 134829274994496 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) I0420 06:41:24.197025 134829274994496 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') I0420 06:41:24.197079 134829274994496 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) I0420 06:41:24.197119 134829274994496 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) I0420 06:41:24.197170 134829274994496 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) I0420 06:41:24.197222 134829274994496 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') I0420 06:41:24.197262 134829274994496 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) I0420 06:41:24.197299 134829274994496 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) I0420 06:41:24.197365 134829274994496 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 06:41:24.197418 134829274994496 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 06:41:24.690720 134829274994496 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:41:24.690815 134829274994496 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 06:41:24.706736 134829274994496 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:41:24.706804 134829274994496 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 06:41:35.482067 134829274994496 max_utils.py:791] Total memory size: 1.3 GB, Output size: 0.1 GB, Temp size: 1.1 GB, Argument size: 0.1 GB, Host temp size: 0.0 GB. I0420 06:41:35.482923 134829274994496 metric_logger.py:301] number parameters: 1.104 billion I0420 06:41:46.753177 134829274994496 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0420 06:41:46.925325 134829274994496 checkpointing.py:776] Waited 0.17213058471679688 seconds for step 0 to finish before starting checkpointing. I0420 06:41:46.927881 134829274994496 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:41:46.929704 134829274994496 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 0 I0420 06:41:46.931285 134829274994496 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints/0. I0420 06:41:47.244657 134829274994496 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 06:41:47.245579 134829274994496 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:41:47.245637 134829274994496 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 I0420 06:41:47.520718 134829274994496 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.276140s I0420 06:41:47.520899 134829274994496 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 5.461 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2824678421020508 s) (per-host) I0420 06:41:47.520950 134829274994496 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.282530s (batch_requests_ready=0.002205s, total_serialization_initiated=0.280250s, others=0.000074s) I0420 06:41:47.521041 134829274994496 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.286638s (all_items=0.000031s, per_item={'items': '0.00003052'}, temp_paths=0.286608) I0420 06:41:47.521870 134829274994496 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.59 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints/0. I0420 06:41:47.522216 134697145329408 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:01:47.522179 I0420 06:41:47.529396 134829274994496 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:41:47.529669 134697103365888 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:41:47.529824 134829274994496 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776667306.927863, 'wait_for_prev_duration_secs': 6.508827209472656e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776667306.9297423, 'checkpointer_blocking_duration_secs': 0.5926408767700195, 'get_old_steps_start_time': 1776667307.5224075, 'get_old_steps_duration_secs': 3.0040740966796875e-05, 'checkpoint_manager_blocking_start_time': 1776667306.9258468, 'checkpoint_manager_blocking_duration_secs': 0.6039397716522217} I0420 06:41:47.529934 134829274994496 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0420 06:41:47.529985 134829274994496 max_utils.py:750] Memstats: After params initialized: I0420 06:41:47.530032 134829274994496 max_utils.py:756] Using (GB) 0.17 / 31.25 (0.544000%) on TPU_16(process=4,(0,4,0,0)) I0420 06:41:47.530064 134829274994496 max_utils.py:756] Using (GB) 0.17 / 31.25 (0.544000%) on TPU_17(process=4,(1,4,0,0)) I0420 06:41:47.530091 134829274994496 max_utils.py:756] Using (GB) 0.17 / 31.25 (0.544000%) on TPU_20(process=4,(0,5,0,0)) I0420 06:41:47.530116 134829274994496 max_utils.py:756] Using (GB) 0.17 / 31.25 (0.544000%) on TPU_21(process=4,(1,5,0,0)) I0420 06:41:47.839903 134829274994496 metric_logger.py:196] completed step: 0, seconds: 11.270, TFLOP/s/device: 1.206, Tokens/s/device: 181.719, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 06:41:48.035623 134829274994496 metric_logger.py:196] completed step: 1, seconds: 1.085, TFLOP/s/device: 12.521, Tokens/s/device: 1887.241, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 06:41:48.445613 134829274994496 metric_logger.py:196] completed step: 2, seconds: 0.033, TFLOP/s/device: 413.246, Tokens/s/device: 62288.999, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0420 06:41:48.608701 134829274994496 metric_logger.py:196] completed step: 3, seconds: 0.410, TFLOP/s/device: 33.141, Tokens/s/device: 4995.439, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.047 I0420 06:41:48.934775 134829274994496 metric_logger.py:196] completed step: 4, seconds: 0.169, TFLOP/s/device: 80.563, Tokens/s/device: 12143.276, total_weights: 65536, loss: 9.278, lm_loss: 9.278, perplexity: 10694.968 I0420 06:41:48.940606 134829274994496 metric_logger.py:196] completed step: 5, seconds: 0.163, TFLOP/s/device: 83.387, Tokens/s/device: 12569.044, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.520 I0420 06:41:51.529401 2739 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 06:41:53.798494 134697111758592 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints/0/items/array_metadatas/process_4 I0420 06:42:10.530746 134829274994496 metric_logger.py:196] completed step: 6, seconds: 0.327, TFLOP/s/device: 41.579, Tokens/s/device: 6267.290, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.688 I0420 06:42:10.693709 134829274994496 metric_logger.py:196] completed step: 7, seconds: 21.428, TFLOP/s/device: 0.634, Tokens/s/device: 95.577, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4377.010 I0420 06:42:10.856554 134829274994496 metric_logger.py:196] completed step: 8, seconds: 0.168, TFLOP/s/device: 81.039, Tokens/s/device: 12215.052, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3845.943 I0420 06:42:11.018957 134829274994496 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0420 06:42:11.019634 134829274994496 checkpointing.py:776] Waited 0.0006961822509765625 seconds for step 9 to finish before starting checkpointing. I0420 06:42:11.021777 134829274994496 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:42:25.800763 134697145329408 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 40.961 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 38.56230092048645 s) (per-host) I0420 06:42:25.800873 134697145329408 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 38.278540s. I0420 06:42:33.538883 134697145329408 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 46.016539s. I0420 06:42:33.539181 134697103365888 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:42:33.539300 134697103365888 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:42:33.539368 134697103365888 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:42:33.541582 134697103365888 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 06:42:33.541779 134829274994496 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0420 06:42:33.541922 134829274994496 checkpoint_manager.py:1452] Waiting for previous save to complete took 22.520144 seconds. If this number is high, consider checkpointing less frequently. I0420 06:42:33.543878 134829274994496 checkpoint_manager.py:1512] [process=4] Saving checkpoint at step 9 I0420 06:42:33.545977 134829274994496 event_tracking.py:70] [process=4] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints/9. I0420 06:42:33.831651 134829274994496 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:42:33.831742 134829274994496 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 I0420 06:42:33.873236 134829274994496 base_pytree_checkpoint_handler.py:154] [process=4][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.042518s I0420 06:42:33.873400 134829274994496 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/blocking_gbytes_per_sec: 33.667 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.045816659927368164 s) (per-host) I0420 06:42:33.873450 134829274994496 base_pytree_checkpoint_handler.py:768] [process=4][thread=MainThread] Initiated Pytree async_save. Time taken: 0.045875s (batch_requests_ready=0.001756s, total_serialization_initiated=0.044053s, others=0.000065s) I0420 06:42:33.873532 134829274994496 composite_checkpoint_handler.py:715] [process=4][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.049973s (all_items=0.000015s, per_item={'items': '0.00001454'}, temp_paths=0.049959) I0420 06:42:33.874163 134829274994496 event_tracking.py:125] [process=4] [async] Finished blocking save in 0.33 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints/9. I0420 06:42:33.874448 134697103365888 async_checkpointer.py:76] [process=4][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 07:02:33.874425 I0420 06:42:33.876346 134829274994496 checkpoint_manager.py:1560] [process=4][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:42:33.876638 134681871120128 async_checkpointer.py:280] [process=4][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:42:33.876798 134829274994496 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776667331.0217478, 'wait_for_prev_duration_secs': 22.520143508911133, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776667353.5439157, 'checkpointer_blocking_duration_secs': 0.3306417465209961, 'get_old_steps_start_time': 1776667353.874583, 'get_old_steps_duration_secs': 2.8848648071289062e-05, 'checkpoint_manager_blocking_start_time': 1776667331.0198786, 'checkpoint_manager_blocking_duration_secs': 22.856884479522705} I0420 06:42:33.876903 134829274994496 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0420 06:42:33.876948 134829274994496 checkpoint_manager.py:2020] [process=4][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:42:38.968666 134697111758592 array_metadata_store.py:203] [process=4][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_11_optimizer_offload_true/checkpoints/9/items/array_metadatas/process_4 I0420 06:43:15.760024 134697103365888 base_pytree_checkpoint_handler.py:130] [process=4] /jax/orbax/write/gbytes_per_sec: 37.669 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.93241548538208 s) (per-host) I0420 06:43:15.760131 134697103365888 async_checkpointer.py:90] [process=4][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.885609s. I0420 06:43:22.221655 134697103365888 async_checkpointer.py:160] [process=4][thread=async_save] Background save thread done. Time taken: 48.347116s. I0420 06:43:22.221959 134681871120128 async_checkpointer.py:288] [process=4][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:43:22.222079 134681871120128 async_checkpointer.py:298] [process=4][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:43:22.222127 134681871120128 checkpoint_manager.py:2137] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:43:22.223710 134681871120128 checkpoint_manager.py:2146] [process=4][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 06:43:22.223901 134829274994496 checkpoint_manager.py:2032] [process=4][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0420 06:43:22.224045 134829274994496 checkpoint_manager.py:2009] [process=4][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:43:22.224955 134829274994496 metric_logger.py:196] completed step: 9, seconds: 0.163, TFLOP/s/device: 83.427, Tokens/s/device: 12574.986, total_weights: 65536, loss: 8.179, lm_loss: 8.179, perplexity: 3564.739 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Mon Apr 20 06:43:31 UTC 2026 EXIT_CODE=0
XPK Start: Mon Apr 20 16:33:31 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. 2026-04-20 16:33:55.733769: 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) I0420 16:33:55.911145 135787014534976 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 16:34:04.951019 135787014534976 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 16:34:04.953417 135787014534976 distributed.py:172] Connecting to JAX distributed service on mt-11-optimizer-offload-84xld-slice-job-0-0.mt-11-optimizer-offload-84xld:8482 I0420 16:34:06.015336 135787014534976 max_utils.py:284] Jax distributed system initialized! I0420 16:34:11.988856 135787014534976 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 16:34:11.988957 135787014534976 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 16:34:11.988996 135787014534976 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 I0420 16:34:11.989027 135787014534976 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 16:34:13.029251 135787014534976 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 16:34:13.029834 135787014534976 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 16:34:13.030013 135787014534976 checkpointing.py:688] Setting up checkpoint logger... I0420 16:34:13.030063 135787014534976 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 16:34:13.030122 135787014534976 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 16:34:13.030455 135787014534976 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 0x7b7ebff1d340>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 16:34:15.721504 135787014534976 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0420 16:34:15.721801 135787014534976 checkpoint_manager.py:708] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b7d9cc92210>}, handler_registry=None I0420 16:34:15.722073 135787014534976 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b7d9cc92210>` 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`. I0420 16:34:15.722144 135787014534976 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b7d9cd7ec00>` 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`. I0420 16:34:15.722183 135787014534976 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 0x7b7d9cc92210>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7b7d9cc92210>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b7d9cd7ec00>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7b7d9cd7ec00>}). I0420 16:34:15.722525 135787014534976 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 16:34:15.722598 135787014534976 async_checkpointer.py:192] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7b6a703e6c00> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 16:34:16.878232 135787014534976 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints I0420 16:34:16.936658 135787014534976 checkpoint_manager.py:929] [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_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_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7b7d9cd7d1c0> I0420 16:34:16.936810 135787014534976 checkpointing.py:302] Checkpoint manager created! I0420 16:34:17.857955 135787014534976 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 16:34:17.858058 135787014534976 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 16:34:18.237999 135787014534976 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 16:34:18.238100 135787014534976 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 16:34:18.254433 135787014534976 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 16:34:18.254492 135787014534976 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 16:34:18.278372 135787014534976 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 16:34:18.278437 135787014534976 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 16:34:18.294673 135787014534976 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 16:34:18.294732 135787014534976 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 16:34:18.310937 135787014534976 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0420 16:34:18.310998 135787014534976 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 16:34:18.335636 135787014534976 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 16:34:18.335702 135787014534976 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 16:34:18.357717 135787014534976 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 16:34:18.357779 135787014534976 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 16:34:18.571098 135787014534976 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0420 16:34:18.571209 135787014534976 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 16:34:20.183461 135787014534976 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 16:34:20.183582 135787014534976 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) I0420 16:34:20.183639 135787014534976 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) I0420 16:34:20.183698 135787014534976 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') I0420 16:34:20.183751 135787014534976 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) I0420 16:34:20.183789 135787014534976 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) I0420 16:34:20.183841 135787014534976 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) I0420 16:34:20.183894 135787014534976 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') I0420 16:34:20.183936 135787014534976 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) I0420 16:34:20.183971 135787014534976 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) I0420 16:34:20.184019 135787014534976 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 16:34:20.184064 135787014534976 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 16:34:20.672625 135787014534976 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 16:34:20.672721 135787014534976 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 16:34:20.687936 135787014534976 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 16:34:20.687996 135787014534976 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 16:34:31.202158 135787014534976 max_utils.py:791] Total memory size: 1.3 GB, Output size: 0.1 GB, Temp size: 1.1 GB, Argument size: 0.1 GB, Host temp size: 0.0 GB. I0420 16:34:31.209977 135787014534976 metric_logger.py:301] number parameters: 1.104 billion I0420 16:34:42.924042 135787014534976 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0420 16:34:43.096274 135787014534976 checkpointing.py:798] Waited 0.172210693359375 seconds for step 0 to finish before starting checkpointing. I0420 16:34:43.098783 135787014534976 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 16:34:43.100500 135787014534976 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 0 I0420 16:34:43.101908 135787014534976 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0. I0420 16:34:43.427315 135787014534976 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 16:34:43.442756 135787014534976 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 16:34:43.442837 135787014534976 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 I0420 16:34:43.709443 135787014534976 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.267426s I0420 16:34:43.710245 135787014534976 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 5.605 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2752854824066162 s) (per-host) I0420 16:34:43.710308 135787014534976 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.275366s (batch_requests_ready=0.002214s, total_serialization_initiated=0.272441s, others=0.000711s) I0420 16:34:43.710384 135787014534976 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.279229s (all_items=0.000019s, per_item={'items': '0.00001860'}, temp_paths=0.279210) I0420 16:34:43.711213 135787014534976 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0. I0420 16:34:43.711507 135640503023360 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 16:54:43.711474 I0420 16:34:43.720538 135787014534976 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 16:34:43.720840 135637877384960 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 16:34:43.720997 135787014534976 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776702883.098764, 'wait_for_prev_duration_secs': 6.031990051269531e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776702883.1005378, 'checkpointer_blocking_duration_secs': 0.611121416091919, 'get_old_steps_start_time': 1776702883.711685, 'get_old_steps_duration_secs': 2.7418136596679688e-05, 'checkpoint_manager_blocking_start_time': 1776702883.0968008, 'checkpoint_manager_blocking_duration_secs': 0.6241588592529297} I0420 16:34:43.721109 135787014534976 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0420 16:34:43.721158 135787014534976 max_utils.py:750] Memstats: After params initialized: I0420 16:34:43.721206 135787014534976 max_utils.py:756] Using (GB) 0.17 / 31.25 (0.544000%) on TPU_0(process=0,(0,0,0,0)) I0420 16:34:43.721237 135787014534976 max_utils.py:756] Using (GB) 0.17 / 31.25 (0.544000%) on TPU_1(process=0,(1,0,0,0)) I0420 16:34:43.721264 135787014534976 max_utils.py:756] Using (GB) 0.17 / 31.25 (0.544000%) on TPU_4(process=0,(0,1,0,0)) I0420 16:34:43.721288 135787014534976 max_utils.py:756] Using (GB) 0.17 / 31.25 (0.544000%) on TPU_5(process=0,(1,1,0,0)) I0420 16:34:43.735839 135640494630656 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0 I0420 16:34:44.066721 135787014534976 metric_logger.py:196] completed step: 0, seconds: 11.400, TFLOP/s/device: 1.192, Tokens/s/device: 179.652, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 16:34:44.069456 135787014534976 metric_logger.py:281] To see full metrics 'tensorboard --logdir=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/tensorboard/' I0420 16:34:44.505887 135787014534976 metric_logger.py:196] completed step: 1, seconds: 1.141, TFLOP/s/device: 11.909, Tokens/s/device: 1795.098, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 16:34:44.667424 135787014534976 metric_logger.py:196] completed step: 2, seconds: 0.440, TFLOP/s/device: 30.892, Tokens/s/device: 4656.334, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0420 16:34:44.830200 135787014534976 metric_logger.py:196] completed step: 3, seconds: 0.009, TFLOP/s/device: 1552.815, Tokens/s/device: 234057.143, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.047 I0420 16:34:45.675710 135640477845248 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0/items I0420 16:34:45.859280 135637833832192 checkpoint.py:188] Wrote Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776702885380316037, 'commit_timestamp_nsecs': None, 'custom_metadata': {}}, json={"item_handlers": null, "metrics": {}, "performance_metrics": {}, "init_timestamp_nsecs": 1776702885380316037, "commit_timestamp_nsecs": null, "custom_metadata": {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0/_CHECKPOINT_METADATA I0420 16:34:46.866862 2793 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 16:34:49.175925 135640452667136 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0/items/array_metadatas/process_0 I0420 16:35:07.626576 135787014534976 metric_logger.py:196] completed step: 4, seconds: 0.163, TFLOP/s/device: 83.613, Tokens/s/device: 12603.077, total_weights: 65536, loss: 9.278, lm_loss: 9.278, perplexity: 10694.968 I0420 16:35:07.634769 135787014534976 metric_logger.py:196] completed step: 5, seconds: 0.163, TFLOP/s/device: 83.402, Tokens/s/device: 12571.281, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.520 I0420 16:35:07.796287 135787014534976 metric_logger.py:196] completed step: 6, seconds: 22.795, TFLOP/s/device: 0.596, Tokens/s/device: 89.845, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.688 I0420 16:35:07.959166 135787014534976 metric_logger.py:196] completed step: 7, seconds: 0.006, TFLOP/s/device: 2203.914, Tokens/s/device: 332197.891, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4377.010 I0420 16:35:08.121757 135787014534976 metric_logger.py:196] completed step: 8, seconds: 0.163, TFLOP/s/device: 83.600, Tokens/s/device: 12601.138, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3845.943 I0420 16:35:08.283760 135787014534976 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0420 16:35:08.284444 135787014534976 checkpointing.py:798] Waited 0.0007033348083496094 seconds for step 9 to finish before starting checkpointing. I0420 16:35:08.286474 135787014534976 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 16:35:24.315817 135639714477824 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 37.457536s (commit=36.193870s, array_metadata_write=1.263666s) I0420 16:35:24.316947 135640503023360 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 38.649 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 40.88195037841797 s) (per-host) I0420 16:35:24.317075 135640503023360 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 40.605454s. I0420 16:35:25.015158 135640503023360 checkpoint.py:228] Read Metadata={'item_handlers': None, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776702885380316037, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} from gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0/_CHECKPOINT_METADATA I0420 16:35:25.792942 135637833832192 checkpoint.py:247] Updated Metadata={'item_handlers': {'items': 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler'}, 'metrics': {}, 'performance_metrics': {}, 'init_timestamp_nsecs': 1776702885380316037, 'commit_timestamp_nsecs': None, 'custom_metadata': {}} to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0/_CHECKPOINT_METADATA I0420 16:35:26.039679 135640503023360 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000259s. I0420 16:35:26.406992 135640503023360 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0420 16:35:27.366956 135640503023360 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.208836s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0/items I0420 16:35:27.368701 135640503023360 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0/items I0420 16:35:27.918113 135640503023360 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0 I0420 16:35:29.783416 135640503023360 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0`. I0420 16:35:29.784165 135640503023360 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 46.68 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/0 I0420 16:35:29.785498 135640503023360 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 46.073873s. I0420 16:35:29.785688 135637877384960 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 16:35:29.785815 135637877384960 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 16:35:29.785881 135637877384960 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 16:35:29.787718 135637877384960 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 16:35:29.787853 135787014534976 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0420 16:35:29.787948 135787014534976 checkpoint_manager.py:1452] Waiting for previous save to complete took 21.501475 seconds. If this number is high, consider checkpointing less frequently. I0420 16:35:29.789824 135787014534976 checkpoint_manager.py:1512] [process=0] Saving checkpoint at step 9 I0420 16:35:29.791963 135787014534976 event_tracking.py:70] [process=0] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9. I0420 16:35:30.106505 135787014534976 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 16:35:30.106604 135787014534976 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 I0420 16:35:30.126322 135787014534976 base_pytree_checkpoint_handler.py:154] [process=0][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.020504s I0420 16:35:30.127167 135787014534976 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/blocking_gbytes_per_sec: 61.765 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.024981975555419922 s) (per-host) I0420 16:35:30.127231 135787014534976 base_pytree_checkpoint_handler.py:768] [process=0][thread=MainThread] Initiated Pytree async_save. Time taken: 0.025061s (batch_requests_ready=0.001889s, total_serialization_initiated=0.022413s, others=0.000759s) I0420 16:35:30.127305 135787014534976 composite_checkpoint_handler.py:715] [process=0][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.028942s (all_items=0.000031s, per_item={'items': '0.00003147'}, temp_paths=0.028911) I0420 16:35:30.127940 135787014534976 event_tracking.py:125] [process=0] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9. I0420 16:35:30.128191 135639793288960 async_checkpointer.py:76] [process=0][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 16:55:30.128168 I0420 16:35:30.138945 135787014534976 checkpoint_manager.py:1560] [process=0][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 16:35:30.139188 135637877384960 async_checkpointer.py:280] [process=0][thread=save_finalize] Waiting for background save thread=async_save. I0420 16:35:30.139317 135787014534976 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776702908.2864437, 'wait_for_prev_duration_secs': 21.50147533416748, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776702929.7898636, 'checkpointer_blocking_duration_secs': 0.33842015266418457, 'get_old_steps_start_time': 1776702930.1283038, 'get_old_steps_duration_secs': 2.8371810913085938e-05, 'checkpoint_manager_blocking_start_time': 1776702908.2846904, 'checkpoint_manager_blocking_duration_secs': 21.854594707489014} I0420 16:35:30.139419 135787014534976 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0420 16:35:30.139462 135787014534976 checkpoint_manager.py:2020] [process=0][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 16:35:30.363008 135640503023360 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9 I0420 16:35:32.147245 135640452667136 atomicity.py:140] Creating tmp directory gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9/items I0420 16:35:35.112384 135639810074368 array_metadata_store.py:203] [process=0][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9/items/array_metadatas/process_0 I0420 16:36:16.559659 135639801681664 base_pytree_checkpoint_handler.py:1282] [process=0][thread=write_metadata_after_commits] Commit + Array metadata written. Time taken: 43.278728s (commit=42.022371s, array_metadata_write=1.256357s) I0420 16:36:16.560755 135639793288960 base_pytree_checkpoint_handler.py:130] [process=0] /jax/orbax/write/gbytes_per_sec: 34.010 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 46.45852541923523 s) (per-host) I0420 16:36:16.560888 135639793288960 async_checkpointer.py:90] [process=0][thread=async_save] 3 Handler Commit operations completed. Time taken: 46.432624s. I0420 16:36:18.139554 135639793288960 array_metadata_store.py:411] [process=0][thread=async_save] Validated ArrayMetadata from all 8 hosts. Time taken: 0.000206s. I0420 16:36:18.925292 135639793288960 ocdbt_utils.py:49] Param validation support for Zarr3 will be added later (b/362328389). I0420 16:36:19.863855 135639793288960 base_pytree_checkpoint_handler.py:1406] [process=0][thread=async_save] Pytree save finalize (merge_ocdbt + ArrayMetadata validation) completed. Time taken: 2.365970s. use_zarr3=True, enable_post_merge_validation=True, directory=gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9/items I0420 16:36:19.869108 135639793288960 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9/items I0420 16:36:20.409371 135639793288960 atomicity.py:666] Finalizing gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9 I0420 16:36:21.863976 135639793288960 atomicity.py:847] [process=0][thread=async_save] Finished saving checkpoint (finalized tmp dir) to `gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9`. I0420 16:36:21.864863 135639793288960 event_tracking.py:138] [process=0] [async] Finished save (blocking + background) in 52.07 seconds @ gs://lance-maxtext/linen_ckpt_xpk_feat_nnx_post_train_fixes_20260420_153543/linen_xpk_feat_nnx_post_train_fixes_20260420_153543_11_optimizer_offload_true/checkpoints/9 I0420 16:36:21.866308 135639793288960 async_checkpointer.py:160] [process=0][thread=async_save] Background save thread done. Time taken: 51.738042s. I0420 16:36:21.866535 135637877384960 async_checkpointer.py:288] [process=0][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 16:36:21.866653 135637877384960 async_checkpointer.py:298] [process=0][thread=save_finalize] No errors found in background save thread=async_save. I0420 16:36:21.866715 135637877384960 checkpoint_manager.py:2137] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 16:36:21.868778 135637877384960 checkpoint_manager.py:2146] [process=0][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 16:36:21.868921 135787014534976 checkpoint_manager.py:2032] [process=0][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0420 16:36:21.869019 135787014534976 checkpoint_manager.py:2009] [process=0][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 16:36:21.870028 135787014534976 metric_logger.py:196] completed step: 9, seconds: 0.163, TFLOP/s/device: 83.425, Tokens/s/device: 12574.677, total_weights: 65536, loss: 8.179, lm_loss: 8.179, perplexity: 3564.739 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Mon Apr 20 16:36:32 UTC 2026 EXIT_CODE=0