main (b117f50cf) vs feat/Migrate-Decoder-And-Tests-to-NNX (4f4617640)| Metric | main b117f50cf | feat/Migrate-Decoder-And-Tests-to-NNX 4f4617640 | Diff (feat/Migrate-Decoder-And-Tests-to-NNX − main) |
|---|---|---|---|
| Parameters | 1.104 billion | 1.104 billion | — |
| Final loss | 8.1790 | 8.1790 | 0 |
| TFLOP/s | 92.139 | 92.098 | -0.041 |
| Tok/s | 13888.2 | 13882.0 | -6.214 |
| Avg s/step | 2.649 | 2.419 | -0.23 |
| Memory % | 1.38 | 1.38 | 0 |
| JAX | 0.9.2 | 0.8.1 | — |
Diff = branch value − main value. Green = branch improved. Red = branch regressed.
XPK Start: Fri Apr 24 07:03:05 UTC 2026 PyTorch was not found. Models won't be available and only tokenizers, configuration and file/data utilities can be used. `rope_parameters`'s factor field must be a float >= 1, got 40 `rope_parameters`'s beta_fast field must be a float, got 32 `rope_parameters`'s beta_slow field must be a float, got 1 DeepseekV32Config got `key=rope_scaling` in kwargs but hasn't set it as attribute. For RoPE standardization you need to set `self.rope_parameters` in model's config. 2026-04-24 07:03:30.657283: 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) I0424 07:03:30.869002 139753873110848 max_utils.py:273] Attempting to initialize the jax distributed system... I0424 07:03:39.908919 139753873110848 distributed.py:149] Starting JAX distributed service on [::]:8482 I0424 07:03:39.911244 139753873110848 distributed.py:172] Connecting to JAX distributed service on mt-01-base-03z8p-slice-job-0-0.mt-01-base-03z8p:8482 I0424 07:03:40.867149 139753873110848 max_utils.py:284] Jax distributed system initialized! I0424 07:03:47.394468 139753873110848 max_utils.py:800] System Information: Jax Version: 0.9.2 I0424 07:03:47.394575 139753873110848 max_utils.py:801] System Information: Jaxlib Version: 0.9.2 I0424 07:03:47.394616 139753873110848 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 I0424 07:03:47.394671 139753873110848 train_utils.py:361] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0424 07:03:48.088459 139753873110848 maxtext_utils.py:1604] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0424 07:03:48.088745 139753873110848 checkpointing.py:677] Setting up checkpoint logger... I0424 07:03:48.088799 139753873110848 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0424 07:03:48.088844 139753873110848 pytree_checkpoint_handler.py:592] save_device_host_concurrent_bytes=None I0424 07:03:48.089180 139753873110848 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 0x7f1a6abe9fd0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0424 07:03:52.038155 139753873110848 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0424 07:03:52.038386 139753873110848 checkpoint_manager.py:708] [process=5][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f060850dc10>}, handler_registry=None I0424 07:03:52.038621 139753873110848 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f060850dc10>` 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`. I0424 07:03:52.038685 139753873110848 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f060866bfb0>` 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`. I0424 07:03:52.038722 139753873110848 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 0x7f060850dc10>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7f060850dc10>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f060866bfb0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7f060866bfb0>}). I0424 07:03:52.039054 139753873110848 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.34 I0424 07:03:52.039124 139753873110848 async_checkpointer.py:192] [process=5][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7f06082ad9e0> timeout: 1200 secs and primary_host=0 for async checkpoint writes I0424 07:03:53.311888 139753873110848 checkpoint_manager.py:1812] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints I0424 07:03:53.319088 139753873110848 checkpoint_manager.py:929] [process=5][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=1, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=FixedIntervalPolicy(interval=10), preservation_policy=LatestN(n=None), prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False, lightweight_initialize=False), root_directory=gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7f060850c1a0> I0424 07:03:53.319210 139753873110848 checkpointing.py:301] Checkpoint manager created! I0424 07:03:54.238256 139753873110848 nnx_wrappers.py:437] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0424 07:03:54.238378 139753873110848 nnx_wrappers.py:437] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0424 07:03:54.619390 139753873110848 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0424 07:03:54.619480 139753873110848 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0424 07:03:54.635864 139753873110848 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0424 07:03:54.635925 139753873110848 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0424 07:03:54.660033 139753873110848 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0424 07:03:54.660105 139753873110848 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 07:03:54.676564 139753873110848 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0424 07:03:54.676623 139753873110848 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 07:03:54.693101 139753873110848 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0424 07:03:54.693160 139753873110848 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 07:03:54.718147 139753873110848 attentions.py:1198] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0424 07:03:54.718216 139753873110848 attentions.py:1198] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 07:03:54.739634 139753873110848 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0424 07:03:54.739719 139753873110848 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0424 07:03:54.948843 139753873110848 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0424 07:03:54.948954 139753873110848 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0424 07:03:56.373929 139753873110848 maxtext_utils.py:1707] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: P('norm',) Physical: (None,) I0424 07:03:56.374059 139753873110848 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0424 07:03:56.374113 139753873110848 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: P('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0424 07:03:56.374172 139753873110848 maxtext_utils.py:1707] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: P('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0424 07:03:56.374224 139753873110848 maxtext_utils.py:1707] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0424 07:03:56.374261 139753873110848 maxtext_utils.py:1707] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: P('norm', 'layers') Physical: (None, None) I0424 07:03:56.374310 139753873110848 maxtext_utils.py:1707] 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) I0424 07:03:56.374359 139753873110848 maxtext_utils.py:1707] 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') I0424 07:03:56.374398 139753873110848 maxtext_utils.py:1707] 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) I0424 07:03:56.374433 139753873110848 maxtext_utils.py:1707] 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) I0424 07:03:56.374478 139753873110848 maxtext_utils.py:1707] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: P('embed_vocab', 'vocab') Physical: ('fsdp', None) I0424 07:03:56.374525 139753873110848 maxtext_utils.py:1707] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: P('vocab', 'embed_vocab') Physical: (None, 'fsdp') I0424 07:03:56.862996 139753873110848 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0424 07:03:56.863083 139753873110848 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0424 07:03:56.878591 139753873110848 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0424 07:03:56.878667 139753873110848 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0424 07:04:07.697839 139753873110848 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. I0424 07:04:07.698580 139753873110848 metric_logger.py:301] number parameters: 1.104 billion I0424 07:04:19.097150 139753873110848 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0424 07:04:19.249784 139753873110848 checkpointing.py:776] Waited 0.15261363983154297 seconds for step 0 to finish before starting checkpointing. I0424 07:04:19.252031 139753873110848 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0424 07:04:19.253644 139753873110848 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 0 I0424 07:04:19.255588 139753873110848 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints/0. I0424 07:04:19.580963 139753873110848 signaling_client.py:364] Using JaxDistributedSignalingClient I0424 07:04:19.581913 139753873110848 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0424 07:04:19.581975 139753873110848 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 I0424 07:04:19.859895 139753873110848 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.279007s I0424 07:04:19.860074 139753873110848 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 5.421 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.28455591201782227 s) (per-host) I0424 07:04:19.860144 139753873110848 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.284632s (batch_requests_ready=0.002249s, total_serialization_initiated=0.282300s, others=0.000083s) I0424 07:04:19.860269 139753873110848 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.288972s (all_items=0.000018s, per_item={'items': '0.00001788'}, temp_paths=0.288954) I0424 07:04:19.862314 139753873110848 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.61 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints/0. I0424 07:04:19.862671 139624907171584 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 07:24:19.862619 I0424 07:04:19.866681 139753873110848 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0424 07:04:19.866984 139623631783680 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0424 07:04:19.867141 139753873110848 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777014259.2520132, 'wait_for_prev_duration_secs': 6.008148193359375e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777014259.253698, 'checkpointer_blocking_duration_secs': 0.6091272830963135, 'get_old_steps_start_time': 1777014259.8628514, 'get_old_steps_duration_secs': 3.0994415283203125e-05, 'checkpoint_manager_blocking_start_time': 1777014259.250202, 'checkpoint_manager_blocking_duration_secs': 0.6168999671936035} I0424 07:04:19.867247 139753873110848 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0424 07:04:19.867300 139753873110848 max_utils.py:750] Memstats: After params initialized: I0424 07:04:19.867347 139753873110848 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_18(process=5,(2,4,0,0)) I0424 07:04:19.867380 139753873110848 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_19(process=5,(3,4,0,0)) I0424 07:04:19.867408 139753873110848 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_22(process=5,(2,5,0,0)) I0424 07:04:19.867434 139753873110848 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_23(process=5,(3,5,0,0)) I0424 07:04:20.176811 139753873110848 metric_logger.py:196] completed step: 0, seconds: 11.398, TFLOP/s/device: 1.192, Tokens/s/device: 179.673, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0424 07:04:20.352935 139753873110848 metric_logger.py:196] completed step: 1, seconds: 1.078, TFLOP/s/device: 12.604, Tokens/s/device: 1899.770, total_weights: 65536, loss: 10.874, lm_loss: 10.874, perplexity: 52779.875 I0424 07:04:20.764144 139753873110848 metric_logger.py:196] completed step: 2, seconds: 0.029, TFLOP/s/device: 466.063, Tokens/s/device: 70250.060, total_weights: 65536, loss: 10.262, lm_loss: 10.262, perplexity: 28637.129 I0424 07:04:20.911367 139753873110848 metric_logger.py:196] completed step: 3, seconds: 0.411, TFLOP/s/device: 33.041, Tokens/s/device: 4980.242, total_weights: 65536, loss: 9.734, lm_loss: 9.734, perplexity: 16889.203 I0424 07:04:21.206738 139753873110848 metric_logger.py:196] completed step: 4, seconds: 0.153, TFLOP/s/device: 88.876, Tokens/s/device: 13396.391, total_weights: 65536, loss: 9.277, lm_loss: 9.277, perplexity: 10694.614 I0424 07:04:21.212270 139753873110848 metric_logger.py:196] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.214, Tokens/s/device: 13899.446, total_weights: 65536, loss: 8.892, lm_loss: 8.892, perplexity: 7270.717 I0424 07:04:22.958622 2840 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0424 07:04:24.965563 139623772800768 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints/0/items/array_metadatas/process_5 I0424 07:04:42.788391 139753873110848 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 I0424 07:04:42.935832 139753873110848 metric_logger.py:196] completed step: 7, seconds: 21.429, TFLOP/s/device: 0.634, Tokens/s/device: 95.569, total_weights: 65536, loss: 8.384, lm_loss: 8.384, perplexity: 4376.905 I0424 07:04:43.083140 139753873110848 metric_logger.py:196] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.374, Tokens/s/device: 13471.469, total_weights: 65536, loss: 8.255, lm_loss: 8.255, perplexity: 3846.406 I0424 07:04:43.229713 139753873110848 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0424 07:04:43.230300 139753873110848 checkpointing.py:776] Waited 0.0006082057952880859 seconds for step 9 to finish before starting checkpointing. I0424 07:04:43.232429 139753873110848 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0424 07:04:57.577780 139624907171584 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 41.566 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 38.00222325325012 s) (per-host) I0424 07:04:57.577902 139624907171584 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 37.715115s. I0424 07:05:04.738688 139624907171584 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 44.875866s. I0424 07:05:04.738976 139623631783680 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0424 07:05:04.739099 139623631783680 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0424 07:05:04.739150 139623631783680 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0424 07:05:04.741672 139623631783680 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0424 07:05:04.741845 139753873110848 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0424 07:05:04.741989 139753873110848 checkpoint_manager.py:1452] Waiting for previous save to complete took 21.509558 seconds. If this number is high, consider checkpointing less frequently. I0424 07:05:04.744000 139753873110848 checkpoint_manager.py:1512] [process=5] Saving checkpoint at step 9 I0424 07:05:04.745966 139753873110848 event_tracking.py:70] [process=5] [async] Started save checkpoint @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints/9. I0424 07:05:05.046016 139753873110848 jax_array_handlers.py:360] Scheduling D2H of 39 prioritized jax.Array. I0424 07:05:05.046113 139753873110848 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 I0424 07:05:05.080097 139753873110848 base_pytree_checkpoint_handler.py:154] [process=5][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.035293s I0424 07:05:05.080270 139753873110848 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/blocking_gbytes_per_sec: 39.763 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.038794517517089844 s) (per-host) I0424 07:05:05.080324 139753873110848 base_pytree_checkpoint_handler.py:768] [process=5][thread=MainThread] Initiated Pytree async_save. Time taken: 0.038858s (batch_requests_ready=0.001811s, total_serialization_initiated=0.036973s, others=0.000074s) I0424 07:05:05.080425 139753873110848 composite_checkpoint_handler.py:715] [process=5][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.043503s (all_items=0.000016s, per_item={'items': '0.00001645'}, temp_paths=0.043486) I0424 07:05:05.081138 139753873110848 event_tracking.py:125] [process=5] [async] Finished blocking save in 0.34 seconds. Continuing save @ gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints/9. I0424 07:05:05.081471 139623631783680 async_checkpointer.py:76] [process=5][thread=async_save] Background save thread started. Deadline for this save operation is 2026-04-24 07:25:05.081433 I0424 07:05:05.083455 139753873110848 checkpoint_manager.py:1560] [process=5][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0424 07:05:05.083796 139603281442560 async_checkpointer.py:280] [process=5][thread=save_finalize] Waiting for background save thread=async_save. I0424 07:05:05.083961 139753873110848 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777014283.2323983, 'wait_for_prev_duration_secs': 21.50955843925476, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777014304.7440405, 'checkpointer_blocking_duration_secs': 0.337566614151001, 'get_old_steps_start_time': 1777014305.081633, 'get_old_steps_duration_secs': 5.14984130859375e-05, 'checkpoint_manager_blocking_start_time': 1777014283.2305334, 'checkpoint_manager_blocking_duration_secs': 21.853387355804443} I0424 07:05:05.084072 139753873110848 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0424 07:05:05.084118 139753873110848 checkpoint_manager.py:2020] [process=5][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0424 07:05:10.379786 139623772800768 array_metadata_store.py:203] [process=5][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_main_20260424_070227/linen_xpk_main_20260424_070227_01_base/checkpoints/9/items/array_metadatas/process_5 I0424 07:05:47.149814 139623631783680 base_pytree_checkpoint_handler.py:130] [process=5] /jax/orbax/write/gbytes_per_sec: 37.513 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.10829472541809 s) (per-host) I0424 07:05:47.149944 139623631783680 async_checkpointer.py:90] [process=5][thread=async_save] 3 Handler Commit operations completed. Time taken: 42.068363s. I0424 07:05:54.428929 139623631783680 async_checkpointer.py:160] [process=5][thread=async_save] Background save thread done. Time taken: 49.347331s. I0424 07:05:54.429230 139603281442560 async_checkpointer.py:288] [process=5][thread=save_finalize] Done with waiting for background save thread=async_save. I0424 07:05:54.429345 139603281442560 async_checkpointer.py:298] [process=5][thread=save_finalize] No errors found in background save thread=async_save. I0424 07:05:54.429395 139603281442560 checkpoint_manager.py:2137] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0424 07:05:54.430986 139603281442560 checkpoint_manager.py:2146] [process=5][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0424 07:05:54.431162 139753873110848 checkpoint_manager.py:2032] [process=5][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0424 07:05:54.431307 139753873110848 checkpoint_manager.py:2009] [process=5][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0424 07:05:54.432125 139753873110848 metric_logger.py:196] completed step: 9, seconds: 0.147, TFLOP/s/device: 92.139, Tokens/s/device: 13888.230, 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: Fri Apr 24 07:06:06 UTC 2026 EXIT_CODE=0
4f4617640
· feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038
· full log
XPK Start: Fri Apr 24 18:01:16 UTC 2026 2026-04-24 18:01:20.844179: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:467] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered WARNING: All log messages before absl::InitializeLog() is called are written to STDERR E0000 00:00:1777053680.857107 10 cuda_dnn.cc:8579] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered E0000 00:00:1777053680.860813 10 cuda_blas.cc:1407] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered W0000 00:00:1777053680.872343 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777053680.872362 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777053680.872365 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. W0000 00:00:1777053680.872367 10 computation_placer.cc:177] computation placer already registered. Please check linkage and avoid linking the same target more than once. 2026-04-24 18:01:40.140790: 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) I0424 18:01:40.660390 136646081435456 max_utils.py:273] Attempting to initialize the jax distributed system... INFO:2026-04-24 18:01:49,701:jax._src.distributed:140: Starting JAX distributed service on [::]:8482 I0424 18:01:49.701055 136646081435456 distributed.py:140] Starting JAX distributed service on [::]:8482 INFO:2026-04-24 18:01:49,703:jax._src.distributed:157: Connecting to JAX distributed service on mt-01-base-vw5lx-slice-job-0-0.mt-01-base-vw5lx:8482 I0424 18:01:49.703393 136646081435456 distributed.py:157] Connecting to JAX distributed service on mt-01-base-vw5lx-slice-job-0-0.mt-01-base-vw5lx:8482 I0424 18:01:58.047182 136646081435456 max_utils.py:284] Jax distributed system initialized! I0424 18:02:05.018219 136646081435456 max_utils.py:800] System Information: Jax Version: 0.8.1 I0424 18:02:05.018327 136646081435456 max_utils.py:801] System Information: Jaxlib Version: 0.8.1 I0424 18:02:05.018368 136646081435456 max_utils.py:802] System Information: Jax Backend: PJRT C API TFRT TPU v6 lite Built on Nov 12 2025 14:16:36 (1762985796) cl/831091709 I0424 18:02:05.018402 136646081435456 train_utils.py:347] WARNING: Sequence packing is essentially ignored for synthetic data. Please use a real dataset to use sequence packing. I0424 18:02:05.636792 136646081435456 maxtext_utils.py:1517] Num_devices: 32, shape (1, 1, 1, 32, 1, 1, 1, 1, 1, 1, 1, 1, 1) I0424 18:02:05.637102 136646081435456 checkpointing.py:677] Setting up checkpoint logger... I0424 18:02:05.637159 136646081435456 checkpointing.py:233] Creating checkpoint manager with ocdbt=True and zarr3=True I0424 18:02:05.637204 136646081435456 pytree_checkpoint_handler.py:589] save_device_host_concurrent_bytes=None I0424 18:02:05.637567 136646081435456 base_pytree_checkpoint_handler.py:415] 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 0x7c46e1578d10>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB) I0424 18:02:09.224718 136646081435456 checkpointing.py:265] Enabling policy for fixed interval checkpointing. I0424 18:02:09.224977 136646081435456 checkpoint_manager.py:709] [process=3][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=('items',), item_handlers={'items': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c346fde9d60>}, handler_registry=None I0424 18:02:09.225211 136646081435456 composite_checkpoint_handler.py:237] Deferred registration for item: "items". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c346fde9d60>` 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`. I0424 18:02:09.225261 136646081435456 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c346f2833e0>` 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`. I0424 18:02:09.225297 136646081435456 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 0x7c346fde9d60>, ('items', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c346fde9d60>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c346f2833e0>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c346f2833e0>}). I0424 18:02:09.225805 136646081435456 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.33 I0424 18:02:09.225885 136646081435456 async_checkpointer.py:177] [process=3][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c346e31eac0> timeout: 600 secs and primary_host=0 for async checkpoint writes I0424 18:02:09.969665 136646081435456 checkpoint_manager.py:1818] Found 0 checkpoint steps in gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints I0424 18:02:10.004512 136646081435456 checkpoint_manager.py:929] [process=3][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_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c346f2825d0> I0424 18:02:10.004653 136646081435456 checkpointing.py:301] Checkpoint manager created! I0424 18:02:10.905240 136646081435456 nnx_wrappers.py:442] Unknown Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_norm_length', 'activation_embed'). I0424 18:02:10.905344 136646081435456 nnx_wrappers.py:442] Unknown Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0424 18:02:11.274170 136646081435456 attentions.py:1088] attentions/inputs_q Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0424 18:02:11.274272 136646081435456 attentions.py:1088] attentions/inputs_q Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0424 18:02:11.290142 136646081435456 attentions.py:1089] attentions/inputs_kv Logical: bfloat16[32,2048,2048]...................................... ('activation_batch', 'activation_attn_length', 'activation_attn_embed'). I0424 18:02:11.290204 136646081435456 attentions.py:1089] attentions/inputs_kv Physical: bfloat16[32,2048,2048]...................................... ('fsdp', None, None). I0424 18:02:11.313248 136646081435456 attentions.py:1154] attentions/query Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0424 18:02:11.313328 136646081435456 attentions.py:1154] attentions/query Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 18:02:11.329315 136646081435456 attentions.py:1155] attentions/key Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0424 18:02:11.329379 136646081435456 attentions.py:1155] attentions/key Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 18:02:11.345301 136646081435456 attentions.py:1156] attentions/value Logical: bfloat16[32,2048,16,128].................................... ('activation_kv_batch', 'activation_attn_length', 'activation_kv_heads', 'activation_kv_head_dim'). I0424 18:02:11.345372 136646081435456 attentions.py:1156] attentions/value Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 18:02:11.370301 136646081435456 attentions.py:1197] attentions/out Logical: bfloat16[32,2048,16,128].................................... ('activation_batch', 'activation_attn_length', 'activation_heads', 'activation_kv'). I0424 18:02:11.370373 136646081435456 attentions.py:1197] attentions/out Physical: bfloat16[32,2048,16,128].................................... ('fsdp', None, None, None). I0424 18:02:11.390003 136646081435456 linears.py:525] linears/x Logical: bfloat16[32,2048,7168]...................................... ('activation_batch', 'activation_length', 'activation_mlp'). I0424 18:02:11.390069 136646081435456 linears.py:525] linears/x Physical: bfloat16[32,2048,7168]...................................... ('fsdp', None, None). I0424 18:02:11.581040 136646081435456 checkpointing.py:577] checkpoint manager exists so trying to load this run's existing checkpoint I0424 18:02:11.581152 136646081435456 checkpointing.py:665] No existing checkpoints found, not restoring checkpoint. fsdp: 32 I0424 18:02:13.024070 136646081435456 maxtext_utils.py:1620] params/params/decoder/decoder_norm/scale Shape: float32[2048] Logical: PartitionSpec('norm',) Physical: (None,) I0424 18:02:13.024222 136646081435456 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_0/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0424 18:02:13.024300 136646081435456 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wi_1/kernel Shape: float32[2048,16,7168] Logical: PartitionSpec('embed', 'layers', 'mlp') Physical: ('fsdp', None, None) I0424 18:02:13.024390 136646081435456 maxtext_utils.py:1620] params/params/decoder/layers/mlp/wo/kernel Shape: float32[7168,16,2048] Logical: PartitionSpec('mlp', 'layers', 'embed') Physical: (None, None, 'fsdp') I0424 18:02:13.024480 136646081435456 maxtext_utils.py:1620] params/params/decoder/layers/post_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0424 18:02:13.024542 136646081435456 maxtext_utils.py:1620] params/params/decoder/layers/pre_self_attention_layer_norm/scale Shape: float32[2048,16] Logical: PartitionSpec('norm', 'layers') Physical: (None, None) I0424 18:02:13.024620 136646081435456 maxtext_utils.py:1620] params/params/decoder/layers/self_attention/key/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0424 18:02:13.024697 136646081435456 maxtext_utils.py:1620] params/params/decoder/layers/self_attention/out/kernel Shape: float32[16,16,128,2048] Logical: PartitionSpec('heads', 'layers', 'kv', 'embed') Physical: (None, None, None, 'fsdp') I0424 18:02:13.024941 136646081435456 maxtext_utils.py:1620] params/params/decoder/layers/self_attention/query/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'q_heads', 'kv') Physical: ('fsdp', None, None, None) I0424 18:02:13.025015 136646081435456 maxtext_utils.py:1620] params/params/decoder/layers/self_attention/value/kernel Shape: float32[2048,16,16,128] Logical: PartitionSpec('embed', 'layers', 'kv_heads', 'kv_head_dim') Physical: ('fsdp', None, None, None) I0424 18:02:13.025102 136646081435456 maxtext_utils.py:1620] params/params/decoder/logits_dense/kernel Shape: float32[2048,32000] Logical: PartitionSpec('embed', 'vocab') Physical: ('fsdp', None) I0424 18:02:13.025185 136646081435456 maxtext_utils.py:1620] params/params/token_embedder/embedding Shape: float32[32000,2048] Logical: PartitionSpec('vocab', 'embed') Physical: (None, 'fsdp') I0424 18:02:13.526132 136646081435456 train.py:155] train/xent Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0424 18:02:13.526236 136646081435456 train.py:155] train/xent Physical: float32[32,2048]............................................ ('fsdp', None). I0424 18:02:13.541324 136646081435456 train.py:162] train/z_loss Logical: float32[32,2048]............................................ ('activation_embed_and_logits_batch', 'activation_length'). I0424 18:02:13.541392 136646081435456 train.py:162] train/z_loss Physical: float32[32,2048]............................................ ('fsdp', None). I0424 18:02:23.933746 136646081435456 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. I0424 18:02:23.934551 136646081435456 metric_logger.py:289] number parameters: 1.104 billion I0424 18:02:34.894861 136646081435456 checkpointing.py:772] Waiting for step 0 to finish before checkpoint... I0424 18:02:35.363524 136646081435456 checkpointing.py:776] Waited 0.4686431884765625 seconds for step 0 to finish before starting checkpointing. I0424 18:02:35.365999 136646081435456 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0424 18:02:35.367846 136646081435456 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 0 I0424 18:02:35.370548 136646081435456 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints/0. I0424 18:02:36.329661 136646081435456 signaling_client.py:364] Using JaxDistributedSignalingClient I0424 18:02:36.330680 136646081435456 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0424 18:02:36.330739 136646081435456 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 I0424 18:02:36.576951 136646081435456 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.247359s I0424 18:02:36.577117 136646081435456 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 3.059 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.504223108291626 s) (per-host) I0424 18:02:36.577171 136646081435456 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.504285s (batch_requests_ready=0.253260s, total_serialization_initiated=0.250957s, others=0.000069s) I0424 18:02:36.577255 136646081435456 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.508581s (all_items=0.000019s, per_item={'items': '0.00001907'}, temp_paths=0.508562) I0424 18:02:36.578199 136527967393536 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0424 18:02:36.578369 136646081435456 async_checkpointer.py:561] Finished blocking save. Time taken: 1.210455s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints/0. I0424 18:02:36.642176 136646081435456 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=0] Starting CheckpointManager Save Finalize thread=save_finalize I0424 18:02:36.642533 136522666669824 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0424 18:02:36.642696 136646081435456 standard_logger.py:34] {'step': 0, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777053755.365982, 'wait_for_prev_duration_secs': 6.079673767089844e-05, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777053755.3678846, 'checkpointer_blocking_duration_secs': 1.2105977535247803, 'get_old_steps_start_time': 1777053756.5785036, 'get_old_steps_duration_secs': 2.574920654296875e-05, 'checkpoint_manager_blocking_start_time': 1777053755.3641386, 'checkpoint_manager_blocking_duration_secs': 1.2785179615020752} I0424 18:02:36.642807 136646081435456 checkpointing.py:408] Started an asynchronous checkpoint save for step 0 I0424 18:02:36.642861 136646081435456 max_utils.py:750] Memstats: After params initialized: I0424 18:02:36.642917 136646081435456 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_10(process=3,(2,2,0,0)) I0424 18:02:36.642967 136646081435456 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_11(process=3,(3,2,0,0)) I0424 18:02:36.642998 136646081435456 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_14(process=3,(2,3,0,0)) I0424 18:02:36.643024 136646081435456 max_utils.py:756] Using (GB) 0.43 / 31.25 (1.376000%) on TPU_15(process=3,(3,3,0,0)) I0424 18:02:36.909018 136646081435456 metric_logger.py:185] completed step: 0, seconds: 10.960, TFLOP/s/device: 1.240, Tokens/s/device: 186.858, total_weights: 65536, loss: 10.874 I0424 18:02:37.086674 136646081435456 metric_logger.py:185] completed step: 1, seconds: 2.013, TFLOP/s/device: 6.750, Tokens/s/device: 1017.469, total_weights: 65536, loss: 10.874 I0424 18:02:37.510282 136646081435456 metric_logger.py:185] completed step: 2, seconds: 0.031, TFLOP/s/device: 443.893, Tokens/s/device: 66908.426, total_weights: 65536, loss: 10.263 I0424 18:02:37.657556 136646081435456 metric_logger.py:185] completed step: 3, seconds: 0.424, TFLOP/s/device: 32.053, Tokens/s/device: 4831.351, total_weights: 65536, loss: 9.734 I0424 18:02:37.952867 136646081435456 metric_logger.py:185] completed step: 4, seconds: 0.153, TFLOP/s/device: 89.011, Tokens/s/device: 13416.663, total_weights: 65536, loss: 9.278 I0424 18:02:37.958534 136646081435456 metric_logger.py:185] completed step: 5, seconds: 0.147, TFLOP/s/device: 92.285, Tokens/s/device: 13910.209, total_weights: 65536, loss: 8.892 I0424 18:02:56.512029 136646081435456 metric_logger.py:185] completed step: 6, seconds: 0.296, TFLOP/s/device: 45.956, Tokens/s/device: 6927.016, total_weights: 65536, loss: 8.592 I0424 18:02:56.659625 136646081435456 metric_logger.py:185] completed step: 7, seconds: 18.407, TFLOP/s/device: 0.738, Tokens/s/device: 111.263, total_weights: 65536, loss: 8.384 I0424 18:02:56.806836 136646081435456 metric_logger.py:185] completed step: 8, seconds: 0.152, TFLOP/s/device: 89.363, Tokens/s/device: 13469.696, total_weights: 65536, loss: 8.255 I0424 18:02:56.954030 136646081435456 checkpointing.py:772] Waiting for step 9 to finish before checkpoint... I0424 18:02:56.954667 136646081435456 checkpointing.py:776] Waited 0.0006592273712158203 seconds for step 9 to finish before starting checkpointing. I0424 18:02:56.956590 136646081435456 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=0][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0424 18:02:57.898883 2670 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com I0424 18:03:00.063392 136527975786240 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints/0/items/array_metadatas/process_3 I0424 18:03:31.710615 136527967393536 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 28.391 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 55.637686252593994 s) (per-host) I0424 18:03:31.710727 136527967393536 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 55.132417s. I0424 18:03:39.060303 136527967393536 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 62.481977s. I0424 18:03:39.060539 136522666669824 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0424 18:03:39.060601 136522666669824 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0424 18:03:39.060680 136522666669824 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is syncing with other hosts... I0424 18:03:39.062359 136522666669824 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=0] CheckpointManager Save Finalize is done on all hosts. I0424 18:03:39.062557 136646081435456 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=0][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=0. W0424 18:03:39.062703 136646081435456 checkpoint_manager.py:1458] Waiting for previous save to complete took 42.106121 seconds. If this number is high, consider checkpointing less frequently. I0424 18:03:39.064775 136646081435456 checkpoint_manager.py:1518] [process=3] Saving checkpoint at step 9 I0424 18:03:39.068096 136646081435456 async_checkpointer.py:452] [process=3] Started async saving checkpoint to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints/9. I0424 18:03:40.037677 136646081435456 jax_array_handlers.py:358] Scheduling D2H of 39 prioritized jax.Array. I0424 18:03:40.037785 136646081435456 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 I0424 18:03:40.081133 136646081435456 base_pytree_checkpoint_handler.py:153] [process=3][thread=MainThread] Initiated "orbax.checkpoint._src.serialization.jax_array_handlers.ArrayHandler".serialize. Time taken: 0.044693s I0424 18:03:40.081302 136646081435456 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/blocking_gbytes_per_sec: 5.103 GiB/s (total gbytes: 1.5 GiB) (time elapsed: 0.30231308937072754 s) (per-host) I0424 18:03:40.081355 136646081435456 base_pytree_checkpoint_handler.py:737] [process=3][thread=MainThread] Initiated Pytree async_save. Time taken: 0.302375s (batch_requests_ready=0.255737s, total_serialization_initiated=0.046568s, others=0.000070s) I0424 18:03:40.081440 136646081435456 composite_checkpoint_handler.py:715] [process=3][thread=MainThread] Initiated CompositeCheckpointHandler.async_save. Time taken: 0.306440s (all_items=0.000017s, per_item={'items': '0.00001669'}, temp_paths=0.306423) I0424 18:03:40.082325 136522675062528 async_checkpointer.py:79] [process=3][thread=async_save] Background save thread started. I0424 18:03:40.082485 136646081435456 async_checkpointer.py:561] Finished blocking save. Time taken: 1.017638s. Continuing background save to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints/9. I0424 18:03:40.087976 136646081435456 checkpoint_manager.py:1566] [process=3][thread=MainThread][step=9] Starting CheckpointManager Save Finalize thread=save_finalize I0424 18:03:40.088175 136522666669824 async_checkpointer.py:265] [process=3][thread=save_finalize] Waiting for background save thread=async_save. I0424 18:03:40.088291 136646081435456 standard_logger.py:34] {'step': 9, 'event_type': 'save', 'directory': 'gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints', 'reached_preemption': False, 'preemption_received_at': None, 'synchronous': False, 'wait_for_prev_start_time': 1777053776.956559, 'wait_for_prev_duration_secs': 42.10612082481384, 'time_between_consecutive_saves_sec': None, 'checkpointer_blocking_start_time': 1777053819.0648146, 'checkpointer_blocking_duration_secs': 1.0177764892578125, 'get_old_steps_start_time': 1777053820.0826087, 'get_old_steps_duration_secs': 2.5510787963867188e-05, 'checkpoint_manager_blocking_start_time': 1777053776.9548888, 'checkpoint_manager_blocking_duration_secs': 43.133368253707886} I0424 18:03:40.088396 136646081435456 checkpointing.py:408] Started an asynchronous checkpoint save for step 9 I0424 18:03:40.088441 136646081435456 checkpoint_manager.py:2024] [process=3][thread=MainThread][step=9][wait_until_finished] Waiting for Save Finalize thread (save_finalize) to complete. I0424 18:03:44.815477 136527975786240 array_metadata_store.py:203] [process=3][thread=array_type_handler] Wrote 39 array_metadata.ArrayMetadata to gs://lance-maxtext/linen_ckpt_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038/linen_xpk_feat_Migrate_Decoder_And_Tests_to_NNX_20260424_180038_01_base/checkpoints/9/items/array_metadatas/process_3 I0424 18:04:21.838260 136522675062528 base_pytree_checkpoint_handler.py:129] [process=3] /jax/checkpoint/write/gbytes_per_sec: 37.556 MiB/s (total gbytes: 1.5 GiB) (time elapsed: 42.0592246055603 s) (per-host) I0424 18:04:21.838380 136522675062528 async_checkpointer.py:90] [process=3][thread=async_save] 3 Handler Commit operations completed. Time taken: 41.755942s. I0424 18:04:28.176989 136522675062528 async_checkpointer.py:144] [process=3][thread=async_save] Background save thread done. Time taken: 48.094535s. I0424 18:04:28.177206 136522666669824 async_checkpointer.py:273] [process=3][thread=save_finalize] Done with waiting for background save thread=async_save. I0424 18:04:28.177264 136522666669824 async_checkpointer.py:283] [process=3][thread=save_finalize] No errors found in background save thread=async_save. I0424 18:04:28.177317 136522666669824 checkpoint_manager.py:2133] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is syncing with other hosts... I0424 18:04:28.179137 136522666669824 checkpoint_manager.py:2142] [process=3][thread=save_finalize][step=9] CheckpointManager Save Finalize is done on all hosts. I0424 18:04:28.179324 136646081435456 checkpoint_manager.py:2036] [process=3][thread=MainThread][step=9][wait_until_finished] Done waiting for Save Finalize thread (save_finalize) running at step=9. I0424 18:04:28.179471 136646081435456 checkpoint_manager.py:2013] [process=3][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning. I0424 18:04:28.180186 136646081435456 metric_logger.py:185] completed step: 9, seconds: 0.148, TFLOP/s/device: 92.098, Tokens/s/device: 13882.016, total_weights: 65536, loss: 8.179 Per train step: Total TFLOPs: 13.59 split as 93.93% learnable weight flops and 6.07% attention flops XPK End: Fri Apr 24 18:04:41 UTC 2026 EXIT_CODE=0