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 | 92.316 | 92.159 | -0.157 |
| Tok/s | 13914.8 | 13891.2 | -23.596 |
| Avg s/step | 2.745 | 2.678 | -0.067 |
| 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: Mon Apr 20 06:33:21 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 06:33:45.840813: 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 06:33:46.023170 138669787989824 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 06:33:55.062855 138669787989824 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 06:33:55.065101 138669787989824 distributed.py:172] Connecting to JAX distributed service on mt-10-shardy-true-bj7e6-slice-job-0-0.mt-10-shardy-true-bj7e6:8482 I0420 06:33:56.348227 138669787989824 max_utils.py:284] Jax distributed system initialized! I0420 06:34:02.595024 138669787989824 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 06:34:02.595125 138669787989824 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 06:34:02.595165 138669787989824 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 06:34:02.595198 138669787989824 train_utils.py:348] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 06:34:03.286667 138669787989824 maxtext_utils.py:1551] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 06:34:03.286953 138669787989824 checkpointing.py:677] Setting up checkpoint logger... I0420 06:34:03.287011 138669787989824 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 06:34:03.287054 138669787989824 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 06:34:03.287373 138669787989824 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 0x7e1dfa1181d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 06:34:06.542274 138669787989824 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0420 06:34:06.542507 138669787989824 checkpoint_manager.py:708] [process=1][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e0a1427ec30>}, handler_registry=None I0420 06:34:06.542752 138669787989824 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e0a1427ec30>` 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 06:34:06.542800 138669787989824 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e0a14074320>` 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 06:34:06.542836 138669787989824 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 0x7e0a1427ec30>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e0a1427ec30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e0a14074320>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e0a14074320>}). I0420 06:34:06.543150 138669787989824 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 06:34:06.543220 138669787989824 async_checkpointer.py:192] [process=1][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e09d456fd80> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 06:34:07.717277 138669787989824 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_10_shardy_true/checkpoints I0420 06:34:07.719482 138669787989824 checkpoint_manager.py:929] [process=1][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_10_shardy_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e0a1426e060> I0420 06:34:07.719591 138669787989824 checkpointing.py:301] Checkpoint manager created! I0420 06:34:08.652229 138669787989824 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 06:34:08.652341 138669787989824 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:34:09.031838 138669787989824 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:34:09.031923 138669787989824 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:34:09.048276 138669787989824 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 06:34:09.048333 138669787989824 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 06:34:09.071806 138669787989824 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 06:34:09.071875 138669787989824 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:34:09.088321 138669787989824 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 06:34:09.088380 138669787989824 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:34:09.104623 138669787989824 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 06:34:09.104679 138669787989824 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:34:09.129299 138669787989824 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 06:34:09.129366 138669787989824 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 06:34:09.151294 138669787989824 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 06:34:09.151360 138669787989824 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 06:34:09.367732 138669787989824 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0420 06:34:09.367836 138669787989824 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 06:34:10.792915 138669787989824 maxtext_utils.py:1654] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 06:34:10.793041 138669787989824 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 06:34:10.793095 138669787989824 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 06:34:10.793152 138669787989824 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 06:34:10.793205 138669787989824 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 06:34:10.793245 138669787989824 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 06:34:10.793297 138669787989824 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 06:34:10.793348 138669787989824 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 06:34:10.793386 138669787989824 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 06:34:10.793421 138669787989824 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 06:34:10.793467 138669787989824 maxtext_utils.py:1654] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 06:34:10.793513 138669787989824 maxtext_utils.py:1654] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 06:34:11.281460 138669787989824 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:34:11.281554 138669787989824 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 06:34:11.296935 138669787989824 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 06:34:11.296997 138669787989824 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 06:34:22.100689 138669787989824 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. I0420 06:34:22.101429 138669787989824 metric_logger.py:301] number parameters: 1.104 billion I0420 06:34:33.186339 138669787989824 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0420 06:34:33.673331 138669787989824 checkpointing.py:776] Waited 0.4869699478149414 seconds for step 0 to finish before starting checkpointing. I0420 06:34:33.675612 138669787989824 checkpoint_manager.py:2009] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:34:33.677211 138669787989824 checkpoint_manager.py:1512] [process=1] Saving checkpoint at step 0 I0420 06:34:33.678628 138669787989824 event_tracking.py:70] [process=1] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_10_shardy_true/checkpoints/0. I0420 06:34:33.965728 138669787989824 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 06:34:33.966835 138669787989824 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:34:33.966891 138669787989824 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 06:34:34.238063 138669787989824 base_pytree_checkpoint_handler.py:154] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.272422s I0420 06:34:34.238229 138669787989824 base_pytree_checkpoint_handler.py:130] [process=1] /jax/orbax/write/blocking_gbytes_per_sec: 5.554 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.27774739265441895 s) (per-host) I0420 06:34:34.238282 138669787989824 base_pytree_checkpoint_handler.py:768] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.277809s (batch_requests_ready=0.002150s, total_serialization_initiated=0.275590s, others=0.000070s) I0420 06:34:34.238375 138669787989824 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.281898s (all_items=0.000017s, per_item={'items': '0.00001693'}, temp_paths=0.281882) I0420 06:34:34.239235 138669787989824 event_tracking.py:125] [process=1] [async] Finished blocking save in 0.56 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_10_shardy_true/checkpoints/0. I0420 06:34:34.239572 138542044251904 async_checkpointer.py:76] [process=1][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:54:34.239534 I0420 06:34:34.258806 138669787989824 checkpoint_manager.py:1560] [process=1][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:34:34.259097 138539886950144 async_checkpointer.py:280] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:34:34.259256 138669787989824 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_10_shardy_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776666873.675594, 'wait_for_prev_duration_secs': 5.9604644775390625e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776666873.677252, 'checkpointer_blocking_duration_secs': 0.5624878406524658, 'get_old_steps_start_time': 1776666874.239765, 'get_old_steps_duration_secs': 2.9802322387695312e-05, 'checkpoint_manager_blocking_start_time': 1776666873.6737976, 'checkpoint_manager_blocking_duration_secs': 0.5854206085205078} I0420 06:34:34.259360 138669787989824 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0420 06:34:34.259409 138669787989824 max_utils.py:750] Memstats: After params initialized: I0420 06:34:34.259459 138669787989824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_2(process=1,(2,0,0,0)) I0420 06:34:34.259490 138669787989824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_3(process=1,(3,0,0,0)) I0420 06:34:34.259518 138669787989824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_6(process=1,(2,1,0,0)) I0420 06:34:34.259543 138669787989824 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_7(process=1,(3,1,0,0)) I0420 06:34:34.570834 138669787989824 metric_logger.py:196] completed step: 0, seconds: 11.085, TFLOP/s/device: 1.226, Tokens/s/device: 184.757, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 06:34:34.739320 138669787989824 metric_logger.py:196] completed step: 1, seconds: 1.383, TFLOP/s/device: 9.825, Tokens/s/device: 1480.884, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 06:34:35.142534 138669787989824 metric_logger.py:196] completed step: 2, seconds: 0.021, TFLOP/s/device: 636.578, Tokens/s/device: 95952.024, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0420 06:34:35.289660 138669787989824 metric_logger.py:196] completed step: 3, seconds: 0.403, TFLOP/s/device: 33.710, Tokens/s/device: 5081.129, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0420 06:34:35.584665 138669787989824 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.732, Tokens/s/device: 13374.694, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0420 06:34:35.590653 138669787989824 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.285, Tokens/s/device: 13910.209, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0420 06:34:37.558171 2845 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 06:34:40.026361 138539895342848 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_10_shardy_true/checkpoints/0/items/array_metadatas/process_1 I0420 06:34:57.741405 138669787989824 metric_logger.py:196] completed step: 6, seconds: 0.295, TFLOP/s/device: 45.982, Tokens/s/device: 6930.884, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0420 06:34:57.888739 138669787989824 metric_logger.py:196] completed step: 7, seconds: 22.004, TFLOP/s/device: 0.617, Tokens/s/device: 93.074, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0420 06:34:58.035913 138669787989824 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.192, Tokens/s/device: 13443.966, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0420 06:34:58.182511 138669787989824 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0420 06:34:58.183194 138669787989824 checkpointing.py:776] Waited 0.0007042884826660156 seconds for step 9 to finish before starting checkpointing. I0420 06:34:58.185218 138669787989824 checkpoint_manager.py:2020] [process=1][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:35:11.577597 138542044251904 base_pytree_checkpoint_handler.py:130] [process=1] /jax/orbax/write/gbytes_per_sec: 41.991 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 37.6170768737793 s) (per-host) I0420 06:35:11.577752 138542044251904 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.338066s. I0420 06:35:19.694545 138542044251904 async_checkpointer.py:160] [process=1][thread=async_save] Background save thread done. Time taken: 45.454843s. I0420 06:35:19.694861 138539886950144 async_checkpointer.py:288] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:35:19.694984 138539886950144 async_checkpointer.py:298] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:35:19.695047 138539886950144 checkpoint_manager.py:2137] [process=1][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:35:19.696367 138539886950144 checkpoint_manager.py:2146] [process=1][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 06:35:19.696532 138669787989824 checkpoint_manager.py:2032] [process=1][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0420 06:35:19.696671 138669787989824 checkpoint_manager.py:1452] Waiting for previous save to complete took 21.511453 seconds. If this number is high, consider checkpointing less frequently. I0420 06:35:19.698804 138669787989824 checkpoint_manager.py:1512] [process=1] Saving checkpoint at step 9 I0420 06:35:19.700759 138669787989824 event_tracking.py:70] [process=1] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_10_shardy_true/checkpoints/9. I0420 06:35:19.986849 138669787989824 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 06:35:19.986942 138669787989824 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 06:35:20.022449 138669787989824 base_pytree_checkpoint_handler.py:154] [process=1][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036659s I0420 06:35:20.022614 138669787989824 base_pytree_checkpoint_handler.py:130] [process=1] /jax/orbax/write/blocking_gbytes_per_sec: 38.509 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.04005718231201172 s) (per-host) I0420 06:35:20.022664 138669787989824 base_pytree_checkpoint_handler.py:768] [process=1][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040117s (batch_requests_ready=0.001756s, total_serialization_initiated=0.038293s, others=0.000068s) I0420 06:35:20.022775 138669787989824 composite_checkpoint_handler.py:715] [process=1][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044333s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.044316) I0420 06:35:20.023467 138669787989824 event_tracking.py:125] [process=1] [async] Finished blocking save in 0.32 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_10_shardy_true/checkpoints/9. I0420 06:35:20.023817 138539886950144 async_checkpointer.py:76] [process=1][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 06:55:20.023780 I0420 06:35:20.028637 138669787989824 checkpoint_manager.py:1560] [process=1][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 06:35:20.028901 138534794090240 async_checkpointer.py:280] [process=1][thread=save_finalize] Waiting for background save thread=async_save. I0420 06:35:20.029072 138669787989824 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_10_shardy_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776666898.1851885, 'wait_for_prev_duration_secs': 21.511452674865723, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776666919.6988437, 'checkpointer_blocking_duration_secs': 0.32512760162353516, 'get_old_steps_start_time': 1776666920.0239944, 'get_old_steps_duration_secs': 3.4332275390625e-05, 'checkpoint_manager_blocking_start_time': 1776666898.1834276, 'checkpoint_manager_blocking_duration_secs': 21.8456072807312} I0420 06:35:20.029187 138669787989824 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0420 06:35:20.029231 138669787989824 checkpoint_manager.py:2020] [process=1][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 06:35:25.394845 138539895342848 array_metadata_store.py:203] [process=1][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260420_060028/linen_xpk_main_20260420_060028_10_shardy_true/checkpoints/9/items/array_metadatas/process_1 I0420 06:36:01.923124 138539886950144 base_pytree_checkpoint_handler.py:130] [process=1] /jax/orbax/write/gbytes_per_sec: 37.663 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 41.94051480293274 s) (per-host) I0420 06:36:01.923249 138539886950144 async_checkpointer.py:90] [process=1][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.899312s. I0420 06:36:10.711187 138539886950144 async_checkpointer.py:160] [process=1][thread=async_save] Background save thread done. Time taken: 50.687236s. I0420 06:36:10.711475 138534794090240 async_checkpointer.py:288] [process=1][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 06:36:10.711595 138534794090240 async_checkpointer.py:298] [process=1][thread=save_finalize] No errors found in background save thread=async_save. I0420 06:36:10.711644 138534794090240 checkpoint_manager.py:2137] [process=1][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 06:36:10.713220 138534794090240 checkpoint_manager.py:2146] [process=1][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 06:36:10.713379 138669787989824 checkpoint_manager.py:2032] [process=1][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0420 06:36:10.713523 138669787989824 checkpoint_manager.py:2009] [process=1][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 06:36:10.714552 138669787989824 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.316, Tokens/s/device: 13914.840, 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: Mon Apr 20 06:36:22 UTC 2026 EXIT_CODE=0
XPK Start: Mon Apr 20 16:21:01 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 16:21:25.022404: 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 16:21:25.200901 136256939337536 max_utils.py:273] Attempting to initialize the jax distributed system... I0420 16:21:34.240396 136256939337536 distributed.py:149] Starting JAX distributed service on [::]:8482 I0420 16:21:34.242785 136256939337536 distributed.py:172] Connecting to JAX distributed service on mt-10-shardy-true-0me1v-slice-job-0-0.mt-10-shardy-true-0me1v:8482 I0420 16:21:36.190870 136256939337536 max_utils.py:284] Jax distributed system initialized! I0420 16:21:42.382565 136256939337536 max_utils.py:800] System Information: Jax Version: 0.9.2 I0420 16:21:42.382663 136256939337536 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0420 16:21:42.382703 136256939337536 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 16:21:42.382741 136256939337536 train_utils.py:378] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0420 16:21:43.554007 136256939337536 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 16:21:43.554591 136256939337536 maxtext_utils.py:1718] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0420 16:21:43.554766 136256939337536 checkpointing.py:688] Setting up checkpoint logger... I0420 16:21:43.554815 136256939337536 checkpointing.py:234] Creating checkpoint manager with ocdbt=True and zarr3=True I0420 16:21:43.554856 136256939337536 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0420 16:21:43.555185 136256939337536 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 0x7bec12a44590>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0420 16:21:46.568031 136256939337536 checkpointing.py:266] Enabling policy for fixed interval checkpointing. I0420 16:21:46.568260 136256939337536 checkpoint_manager.py:708] [process=2][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7beb0668d6d0>}, handler_registry=None I0420 16:21:46.568493 136256939337536 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7beb0668d6d0>` 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 16:21:46.568553 136256939337536 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7beb0669fd40>` 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 16:21:46.568589 136256939337536 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 0x7beb0668d6d0>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7beb0668d6d0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7beb0669fd40>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7beb0669fd40>}). I0420 16:21:46.568911 136256939337536 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0420 16:21:46.568978 136256939337536 async_checkpointer.py:192] [process=2][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7bd6d856ec00> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0420 16:21:47.265985 136256939337536 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_10_shardy_true/checkpoints I0420 16:21:47.720731 136256939337536 checkpoint_manager.py:929] [process=2][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_10_shardy_true/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7beb066cb740> I0420 16:21:47.720898 136256939337536 checkpointing.py:302] Checkpoint manager created! I0420 16:21:48.662637 136256939337536 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0420 16:21:48.662748 136256939337536 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 16:21:49.044250 136256939337536 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 16:21:49.044341 136256939337536 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 16:21:49.060410 136256939337536 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0420 16:21:49.060469 136256939337536 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0420 16:21:49.083897 136256939337536 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 16:21:49.083966 136256939337536 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 16:21:49.100179 136256939337536 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 16:21:49.100238 136256939337536 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 16:21:49.116633 136256939337536 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 16:21:49.116693 136256939337536 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 16:21:49.140953 136256939337536 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0420 16:21:49.141026 136256939337536 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0420 16:21:49.162504 136256939337536 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0420 16:21:49.162586 136256939337536 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0420 16:21:49.375737 136256939337536 checkpointing.py:578] checkpoint manager exists so trying to load this run's existing checkpoint I0420 16:21:49.375842 136256939337536 checkpointing.py:676] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0420 16:21:50.791969 136256939337536 maxtext_utils.py:1827] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0420 16:21:50.792091 136256939337536 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 16:21:50.792155 136256939337536 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 16:21:50.792237 136256939337536 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 16:21:50.792296 136256939337536 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 16:21:50.792340 136256939337536 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 16:21:50.792391 136256939337536 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 16:21:50.792442 136256939337536 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 16:21:50.792482 136256939337536 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 16:21:50.792578 136256939337536 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 16:21:50.792720 136256939337536 maxtext_utils.py:1827] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0420 16:21:50.792820 136256939337536 maxtext_utils.py:1827] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0420 16:21:51.279277 136256939337536 train.py:157] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 16:21:51.279367 136256939337536 train.py:157] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0420 16:21:51.294624 136256939337536 train.py:164] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0420 16:21:51.294683 136256939337536 train.py:164] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0420 16:22:02.187316 136256939337536 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. I0420 16:22:02.188121 136256939337536 metric_logger.py:301] number parameters: 1.104 billion I0420 16:22:13.574791 136256939337536 checkpointing.py:794] Waiting for step 0 to finish before checkpoint... I0420 16:22:13.729152 136256939337536 checkpointing.py:798] Waited 0.154341459274292 seconds for step 0 to finish before starting checkpointing. I0420 16:22:13.731824 136256939337536 checkpoint_manager.py:2009] [process=2][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 16:22:13.733484 136256939337536 checkpoint_manager.py:1512] [process=2] Saving checkpoint at step 0 I0420 16:22:13.734972 136256939337536 event_tracking.py:70] [process=2] [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_10_shardy_true/checkpoints/0. I0420 16:22:14.522557 136256939337536 signaling_client.py:364] Using JaxDistributedSignalingClient I0420 16:22:14.523475 136256939337536 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 16:22:14.523551 136256939337536 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 16:22:14.794772 136256939337536 base_pytree_checkpoint_handler.py:154] [process=2][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.272307s I0420 16:22:14.794940 136256939337536 base_pytree_checkpoint_handler.py:130] [process=2] /jax/orbax/write/blocking_gbytes_per_sec: 5.556 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.2776374816894531 s) (per-host) I0420 16:22:14.795001 136256939337536 base_pytree_checkpoint_handler.py:768] [process=2][thread=MainThread] Initiated Pytree async_save. Time taken: 0.277705s (batch_requests_ready=0.002147s, total_serialization_initiated=0.275483s, others=0.000075s) I0420 16:22:14.795126 136256939337536 composite_checkpoint_handler.py:715] [process=2][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.281895s (all_items=0.000017s, per_item={'items': '0.00001717'}, temp_paths=0.281878) I0420 16:22:14.796045 136256939337536 event_tracking.py:125] [process=2] [async] Finished blocking save in 1.06 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_10_shardy_true/checkpoints/0. I0420 16:22:14.796322 136127973418752 async_checkpointer.py:76] [process=2][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 16:42:14.796289 I0420 16:22:14.807702 136256939337536 checkpoint_manager.py:1560] [process=2][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0420 16:22:14.808000 136126314313472 async_checkpointer.py:280] [process=2][thread=save_finalize] Waiting for background save thread=async_save. I0420 16:22:14.808172 136256939337536 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_10_shardy_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776702133.7318063, 'wait_for_prev_duration_secs': 6.0558319091796875e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776702133.7335331, 'checkpointer_blocking_duration_secs': 1.0629374980926514, 'get_old_steps_start_time': 1776702134.796494, 'get_old_steps_duration_secs': 3.981590270996094e-05, 'checkpoint_manager_blocking_start_time': 1776702133.729719, 'checkpoint_manager_blocking_duration_secs': 1.0784144401550293} I0420 16:22:14.808295 136256939337536 checkpointing.py:409] Started an asynchronous checkpoint save for step 0 I0420 16:22:14.808352 136256939337536 max_utils.py:750] Memstats: After params initialized: I0420 16:22:14.808405 136256939337536 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_8(process=2,(0,2,0,0)) I0420 16:22:14.808438 136256939337536 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_9(process=2,(1,2,0,0)) I0420 16:22:14.808476 136256939337536 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_12(process=2,(0,3,0,0)) I0420 16:22:14.808499 136256939337536 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_13(process=2,(1,3,0,0)) I0420 16:22:15.121955 136256939337536 metric_logger.py:196] completed step: 0, seconds: 11.387, TFLOP/s/device: 1.193, Tokens/s/device: 179.861, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 16:22:15.300716 136256939337536 metric_logger.py:196] completed step: 1, seconds: 1.546, TFLOP/s/device: 8.791, Tokens/s/device: 1325.025, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0420 16:22:15.722165 136256939337536 metric_logger.py:196] completed step: 2, seconds: 0.032, TFLOP/s/device: 429.456, Tokens/s/device: 64732.284, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0420 16:22:15.869572 136256939337536 metric_logger.py:196] completed step: 3, seconds: 0.422, TFLOP/s/device: 32.211, Tokens/s/device: 4855.175, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0420 16:22:16.164486 136256939337536 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.006, Tokens/s/device: 13415.960, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0420 16:22:16.170174 136256939337536 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.259, Tokens/s/device: 13906.336, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0420 16:22:17.937968 2826 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0420 16:22:20.330994 136126322706176 array_metadata_store.py:203] [process=2][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_10_shardy_true/checkpoints/0/items/array_metadatas/process_2 I0420 16:22:37.527109 136256939337536 metric_logger.py:196] completed step: 6, seconds: 0.295, TFLOP/s/device: 46.005, Tokens/s/device: 6934.428, total_weights: 65536, loss: 8.592, lm_loss: 8.592, perplexity: 5390.759 I0420 16:22:37.674643 136256939337536 metric_logger.py:196] completed step: 7, seconds: 21.211, TFLOP/s/device: 0.641, Tokens/s/device: 96.556, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0420 16:22:37.821959 136256939337536 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.493, Tokens/s/device: 13489.303, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0420 16:22:37.968419 136256939337536 checkpointing.py:794] Waiting for step 9 to finish before checkpoint... I0420 16:22:37.969089 136256939337536 checkpointing.py:798] Waited 0.000690460205078125 seconds for step 9 to finish before starting checkpointing. I0420 16:22:37.971122 136256939337536 checkpoint_manager.py:2020] [process=2][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 16:22:49.988355 136127973418752 base_pytree_checkpoint_handler.py:130] [process=2] /jax/orbax/write/gbytes_per_sec: 44.532 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 35.47101449966431 s) (per-host) I0420 16:22:49.988481 136127973418752 async_checkpointer.py:90] [process=2][thread=async_save] 3 Handler Commit operations completed. Time taken: 35.192050s. I0420 16:22:59.863914 136127973418752 async_checkpointer.py:160] [process=2][thread=async_save] Background save thread done. Time taken: 45.067466s. I0420 16:22:59.864187 136126314313472 async_checkpointer.py:288] [process=2][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 16:22:59.864306 136126314313472 async_checkpointer.py:298] [process=2][thread=save_finalize] No errors found in background save thread=async_save. I0420 16:22:59.864356 136126314313472 checkpoint_manager.py:2137] [process=2][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0420 16:22:59.866232 136126314313472 checkpoint_manager.py:2146] [process=2][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0420 16:22:59.866433 136256939337536 checkpoint_manager.py:2032] [process=2][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0420 16:22:59.866586 136256939337536 checkpoint_manager.py:1452] Waiting for previous save to complete took 21.895463 seconds. If this number is high, consider checkpointing less frequently. I0420 16:22:59.868420 136256939337536 checkpoint_manager.py:1512] [process=2] Saving checkpoint at step 9 I0420 16:22:59.870510 136256939337536 event_tracking.py:70] [process=2] [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_10_shardy_true/checkpoints/9. I0420 16:23:00.600843 136256939337536 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0420 16:23:00.600931 136256939337536 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 16:23:00.636510 136256939337536 base_pytree_checkpoint_handler.py:154] [process=2][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.036779s I0420 16:23:00.636677 136256939337536 base_pytree_checkpoint_handler.py:130] [process=2] /jax/orbax/write/blocking_gbytes_per_sec: 38.447 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.040122270584106445 s) (per-host) I0420 16:23:00.636726 136256939337536 base_pytree_checkpoint_handler.py:768] [process=2][thread=MainThread] Initiated Pytree async_save. Time taken: 0.040181s (batch_requests_ready=0.001705s, total_serialization_initiated=0.038411s, others=0.000065s) I0420 16:23:00.636813 136256939337536 composite_checkpoint_handler.py:715] [process=2][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.044410s (all_items=0.000016s, per_item={'items': '0.00001621'}, temp_paths=0.044394) I0420 16:23:00.637490 136256939337536 event_tracking.py:125] [process=2] [async] Finished blocking save in 0.77 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_10_shardy_true/checkpoints/9. I0420 16:23:00.637793 136126314313472 async_checkpointer.py:76] [process=2][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-20 16:43:00.637759 I0420 16:23:00.641194 136256939337536 checkpoint_manager.py:1560] [process=2][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0420 16:23:00.641424 136103453505280 async_checkpointer.py:280] [process=2][thread=save_finalize] Waiting for background save thread=async_save. I0420 16:23:00.641562 136256939337536 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_10_shardy_true/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1776702157.9710941, 'wait_for_prev_duration_secs': 21.895463228225708, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1776702179.868458, 'checkpointer_blocking_duration_secs': 0.7694671154022217, 'get_old_steps_start_time': 1776702180.6379488, 'get_old_steps_duration_secs': 2.8371810913085938e-05, 'checkpoint_manager_blocking_start_time': 1776702157.9693122, 'checkpoint_manager_blocking_duration_secs': 22.672214031219482} I0420 16:23:00.641669 136256939337536 checkpointing.py:409] Started an asynchronous checkpoint save for step 9 I0420 16:23:00.641715 136256939337536 checkpoint_manager.py:2020] [process=2][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0420 16:23:06.244881 136126322706176 array_metadata_store.py:203] [process=2][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_10_shardy_true/checkpoints/9/items/array_metadatas/process_2 I0420 16:23:43.103858 136126314313472 base_pytree_checkpoint_handler.py:130] [process=2] /jax/orbax/write/gbytes_per_sec: 37.161 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.50726556777954 s) (per-host) I0420 16:23:43.103983 136126314313472 async_checkpointer.py:90] [process=2][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.466084s. I0420 16:23:51.982349 136126314313472 async_checkpointer.py:160] [process=2][thread=async_save] Background save thread done. Time taken: 51.344436s. I0420 16:23:51.982595 136103453505280 async_checkpointer.py:288] [process=2][thread=save_finalize] Done with waiting for background save thread=async_save. I0420 16:23:51.982658 136103453505280 async_checkpointer.py:298] [process=2][thread=save_finalize] No errors found in background save thread=async_save. I0420 16:23:51.982704 136103453505280 checkpoint_manager.py:2137] [process=2][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0420 16:23:51.985495 136103453505280 checkpoint_manager.py:2146] [process=2][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0420 16:23:51.985674 136256939337536 checkpoint_manager.py:2032] [process=2][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0420 16:23:51.985826 136256939337536 checkpoint_manager.py:2009] [process=2][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0420 16:23:51.986768 136256939337536 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.159, Tokens/s/device: 13891.244, 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: Mon Apr 20 16:24:01 UTC 2026 EXIT_CODE=0