XPK Start: Mon Apr 20 19:32:02 UTC 2026
2026-04-20 19:32:31.321086: 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 19:32:31.544836 138986743772992 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-20 19:32:40,585:jax._src.distributed:149: Starting JAX distributed service on [::]:8482
I0420 19:32:40.585738 138986743772992 distributed.py:149] Starting JAX distributed service on [::]:8482
INFO:2026-04-20 19:32:40,588:jax._src.distributed:166: Connecting to JAX distributed service on mt-01-sft-smoke-g940r-slice-job-0-0.mt-01-sft-smoke-g940r:8482
I0420 19:32:40.588216 138986743772992 distributed.py:166] Connecting to JAX distributed service on mt-01-sft-smoke-g940r-slice-job-0-0.mt-01-sft-smoke-g940r:8482
I0420 19:32:41.982975 138986743772992 max_utils.py:284] Jax distributed system initialized!
I0420 19:32:48.349796 138986743772992 max_utils.py:800] System Information: Jax Version: 0.8.3
I0420 19:32:48.349907 138986743772992 max_utils.py:801] System Information: Jaxlib Version: 0.8.3
I0420 19:32:48.349949 138986743772992 max_utils.py:802] System Information: Jax Backend: PJRT C API
TFRT TPU v6 lite
Built on Dec 15 2025 14:03:46 (1765836226) cl/844590465
I0420 19:32:48.353533 138986743772992 maxtext_utils.py:1631] Num_devices: 32, shape (1, 4, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0420 19:32:48.448008 138986743772992 maxtext_utils.py:1631] Num_devices: 32, shape (1, 4, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0420 19:32:48.547594 138986743772992 maxtext_utils.py:1631] Num_devices: 32, shape (1, 4, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0420 19:32:49.675107 138986743772992 config.py:112] TensorFlow version 2.20.0 available.
I0420 19:32:49.675602 138986743772992 config.py:125] JAX version 0.8.3 available.
/deps/src/maxtext/input_pipeline/input_pipeline_utils.py:467: UserWarning: WARNING: Inefficient dataloading. Your train or eval dataset contains 3 shards, smaller than number of host loading data. This is known to lead to inefficient dataloading. Seegithub.com/google/maxtext/blob/main/getting_started/Data_Input_Pipeline.md#multihost-dataloading-best-practice
warnings.warn(
E0420 19:32:55.342401 138986743772992 packing.py:209] PackAndBatchOperation is deprecated. Please use lazy_dataset.FirstFitPackIterDataset instead.
I0420 19:32:55.342613 138986743772992 data_loader.py:408] Adding CopyNumPyArrayToSharedMemory MapTransform.
I0420 19:32:55.674550 138986743772992 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0420 19:32:55.674990 138986743772992 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7e67a9d865a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 19:32:55.675039 138986743772992 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0420 19:32:55.675094 138986743772992 base_pytree_checkpoint_handler.py:411] Created BasePyTreeCheckpointHandler: use_ocdbt=True, use_zarr3=False, pytree_metadata_options=PyTreeMetadataOptions(support_rich_types=False), array_metadata_store=<orbax.checkpoint._src.metadata.array_metadata_store.Store object at 0x7e67a9d865a0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0420 19:32:55.675144 138986743772992 checkpoint_manager.py:702] [process=0][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=None, item_handlers={'model_params': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e5d0468f230>, 'optimizer_state': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e4a481101a0>, 'custom_metadata': <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4a48111c40>}, handler_registry=None
I0420 19:32:55.675396 138986743772992 composite_checkpoint_handler.py:237] Deferred registration for item: "model_params". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e5d0468f230>` for item "model_params" 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 19:32:55.675444 138986743772992 composite_checkpoint_handler.py:237] Deferred registration for item: "optimizer_state". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e4a481101a0>` for item "optimizer_state" 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 19:32:55.675474 138986743772992 composite_checkpoint_handler.py:237] Deferred registration for item: "custom_metadata". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4a48111c40>` for item "custom_metadata" 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 19:32:55.675500 138986743772992 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4a50131970>` 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 19:32:55.675529 138986743772992 composite_checkpoint_handler.py:505] Initialized registry DefaultCheckpointHandlerRegistry({('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e5d0468f230>, ('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e5d0468f230>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e4a481101a0>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7e4a481101a0>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4a48111c40>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4a48111c40>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4a50131970>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7e4a50131970>}).
I0420 19:32:55.675722 138986743772992 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
I0420 19:32:55.675772 138986743772992 async_checkpointer.py:177] [process=0][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7e4a50032660> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0420 19:32:57.701300 138986743772992 checkpoint_manager.py:558] Created directory=gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190413/pt_sft_nnx_xpk_feat_nnx_set_defaults_true_20260420_190413_01_sft_smoke/checkpoints
I0420 19:32:58.134330 138986743772992 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190413/pt_sft_nnx_xpk_feat_nnx_set_defaults_true_20260420_190413_01_sft_smoke/checkpoints
I0420 19:32:58.151312 138986743772992 checkpoint_manager.py:921] [process=0][thread=MainThread] CheckpointManager created, primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=10000, max_to_keep=None, keep_time_interval=None, keep_period=None, should_keep_fn=None, best_fn=None, best_mode='max', keep_checkpoints_without_metrics=True, step_prefix=None, step_format_fixed_length=None, step_name_format=None, create=True, cleanup_tmp_directories=False, save_on_steps=frozenset(), single_host_load_and_broadcast=False, todelete_subdir=None, todelete_full_path=None, enable_hns=False, enable_background_delete=False, read_only=False, enable_async_checkpointing=True, async_options=None, multiprocessing_options=MultiprocessingOptions(primary_host=0, active_processes=None, barrier_sync_key_prefix=None), should_save_fn=None, file_options=FileOptions(path_permission_mode=None), save_root_metadata=True, temporary_path_class=None, save_decision_policy=None, preservation_policy=None, prevent_write_metrics=False, enable_should_save_is_saving_in_progress_check=True, enable_per_process_directory_creation=False), root_directory=gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_set_defaults_true_20260420_190413/pt_sft_nnx_xpk_feat_nnx_set_defaults_true_20260420_190413_01_sft_smoke/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7e4a481113d0>
I0420 19:32:58.156954 138986743772992 metrics_logger.py:64] WandbBackend skipped: 'wandb' library not installed.
I0420 19:32:58.157164 138986743772992 peft_trainer.py:590] Training with mesh: Mesh('diloco': 1, 'data': 4, 'stage': 1, 'fsdp': 8, 'fsdp_transpose': 1, 'sequence': 1, 'context': 1, 'context_autoregressive': 1, 'tensor': 1, 'tensor_transpose': 1, 'tensor_sequence': 1, 'expert': 1, 'autoregressive': 1, axis_types=(Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto, Auto))
I0420 19:32:58.575567 138986743772992 peft_trainer.py:600] Compiled train_step cache size: 0
Training: 0%| | 0/5 [00:00<?, ?step/s]I0420 19:32:58.580372 138986743772992 metric_logger.py:289] number parameters: 0.000 billion
I0420 19:32:58.645378 138833710528256 grain_pool.py:367] Grain pool will use 1 processes.
I0420 19:32:58.671885 138833710528256 grain_pool.py:440] Grain pool will start child processes.
Per train step:
Total TFLOPs: 0.00
split as 54.29% learnable weight flops and 45.71% attention flops
I0420 19:32:58.677323 138833710528256 grain_pool.py:448] Grain pool started all child processes.
2026-04-20 19:33:02.583099: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2026-04-20 19:33:02.627836: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI AVX512_BF16 FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2026-04-20 19:33:03.609565: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2026-04-20 19:33:07.828553: 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 19:33:13.722692 138986743772992 utils.py:86] Train loop finished in: 15.0787 seconds
Traceback (most recent call last):
File "<frozen runpy>", line 198, in _run_module_as_main
File "<frozen runpy>", line 88, in _run_code
File "/deps/src/maxtext/trainers/post_train/sft/train_sft.py", line 216, in <module>
app.run(main)
File "/usr/local/lib/python3.12/site-packages/absl/app.py", line 316, in run
_run_main(main, args)
File "/usr/local/lib/python3.12/site-packages/absl/app.py", line 261, in _run_main
sys.exit(main(argv))
^^^^^^^^^^
File "/deps/src/maxtext/trainers/post_train/sft/train_sft.py", line 212, in main
train(mt_config, goodput_recorder)
File "/deps/src/maxtext/trainers/post_train/sft/train_sft.py", line 189, in train
trainer = train_model(mt_config, trainer, mesh)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/deps/src/maxtext/trainers/post_train/sft/train_sft.py", line 175, in train_model
trainer.train(trainer.data_hooks.train_data_iterator, trainer.data_hooks.eval_data_iterator)
File "/usr/local/lib/python3.12/site-packages/tunix/sft/peft_trainer.py", line 659, in train
train_example = sharding_utils.shard_input(
^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/tunix/sft/sharding_utils.py", line 58, in shard_input
return jax.tree.map(
^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/tree.py", line 155, in map
return tree_util.tree_map(f, tree, *rest, is_leaf=is_leaf)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/tree_util.py", line 362, in tree_map
return treedef.unflatten(f(*xs) for xs in zip(*all_leaves))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/tree_util.py", line 362, in <genexpr>
return treedef.unflatten(f(*xs) for xs in zip(*all_leaves))
^^^^^^
File "/usr/local/lib/python3.12/site-packages/tunix/sft/sharding_utils.py", line 59, in <lambda>
lambda x: jax.make_array_from_process_local_data(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/array.py", line 986, in make_array_from_process_local_data
out = [_array_from_process_local_data(data, s, shape)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/array.py", line 1048, in _array_from_process_local_data
return make_array_from_callback(global_shape, sharding, cb)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/array.py", line 845, in make_array_from_callback
per_device_values = api.device_put(per_device_values, devices)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/api.py", line 2729, in device_put
out_flat = dispatch._batched_device_put_impl(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/dispatch.py", line 558, in _batched_device_put_impl
y = _device_put_impl(x, device=device, src=src, copy=cp, aval=aval)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/dispatch.py", line 545, in _device_put_impl
return _device_put_sharding_impl(x, aval, device, copy)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/jax/_src/dispatch.py", line 487, in _device_put_sharding_impl
raise ValueError(
ValueError: device_put's first argument must be a fully addressable array, but got value with devices {TpuDevice(id=8, process_index=2, coords=(0,2,0), core_on_chip=0), TpuDevice(id=3, process_index=1, coords=(3,0,0), core_on_chip=0), TpuDevice(id=20, process_index=4, coords=(0,5,0), core_on_chip=0), TpuDevice(id=28, process_index=6, coords=(0,7,0), core_on_chip=0), TpuDevice(id=6, process_index=1, coords=(2,1,0), core_on_chip=0), TpuDevice(id=22, process_index=5, coords=(2,5,0), core_on_chip=0), TpuDevice(id=4, process_index=0, coords=(0,1,0), core_on_chip=0), TpuDevice(id=31, process_index=7, coords=(3,7,0), core_on_chip=0), TpuDevice(id=10, process_index=3, coords=(2,2,0), core_on_chip=0), TpuDevice(id=15, process_index=3, coords=(3,3,0), core_on_chip=0), TpuDevice(id=21, process_index=4, coords=(1,5,0), core_on_chip=0), TpuDevice(id=29, process_index=6, coords=(1,7,0), core_on_chip=0), TpuDevice(id=11, process_index=3, coords=(3,2,0), core_on_chip=0), TpuDevice(id=23, process_index=5, coords=(3,5,0), core_on_chip=0), TpuDevice(id=7, process_index=1, coords=(3,1,0), core_on_chip=0), TpuDevice(id=9, process_index=2, coords=(1,2,0), core_on_chip=0), TpuDevice(id=1, process_index=0, coords=(1,0,0), core_on_chip=0), TpuDevice(id=16, process_index=4, coords=(0,4,0), core_on_chip=0), TpuDevice(id=24, process_index=6, coords=(0,6,0), core_on_chip=0), TpuDevice(id=18, process_index=5, coords=(2,4,0), core_on_chip=0), TpuDevice(id=26, process_index=7, coords=(2,6,0), core_on_chip=0), TpuDevice(id=12, process_index=2, coords=(0,3,0), core_on_chip=0), TpuDevice(id=2, process_index=1, coords=(2,0,0), core_on_chip=0), TpuDevice(id=0, process_index=0, coords=(0,0,0), core_on_chip=0), TpuDevice(id=17, process_index=4, coords=(1,4,0), core_on_chip=0), TpuDevice(id=25, process_index=6, coords=(1,6,0), core_on_chip=0), TpuDevice(id=19, process_index=5, coords=(3,4,0), core_on_chip=0), TpuDevice(id=27, process_index=7, coords=(3,6,0), core_on_chip=0), TpuDevice(id=5, process_index=0, coords=(1,1,0), core_on_chip=0), TpuDevice(id=13, process_index=2, coords=(1,3,0), core_on_chip=0), TpuDevice(id=14, process_index=3, coords=(2,3,0), core_on_chip=0), TpuDevice(id=30, process_index=7, coords=(2,7,0), core_on_chip=0)}
I0420 19:33:14.073292 138833710528256 grain_pool.py:542] Grain pool is exiting.
I0420 19:33:14.073394 138833710528256 grain_pool.py:547] Shutting down multiprocessing system.
I0420 19:33:19.914500 138833710528256 grain_pool.py:547] Shutting down multiprocessing system.
Training: 0%| | 0/5 [00:24<?, ?step/s]
Exception ignored in: <function GCSRecordWriter.__del__ at 0x7e67a3b05440>
Traceback (most recent call last):
File "/usr/local/lib/python3.12/site-packages/tensorboardX/record_writer.py", line 134, in __del__
File "/usr/local/lib/python3.12/site-packages/tensorboardX/record_writer.py", line 158, in close
File "/usr/local/lib/python3.12/site-packages/tensorboardX/record_writer.py", line 149, in flush
File "/usr/local/lib/python3.12/copy.py", line 87, in copy
ImportError: sys.meta_path is None, Python is likely shutting down
/usr/local/lib/python3.12/multiprocessing/resource_tracker.py:279: UserWarning: resource_tracker: There appear to be 15 leaked shared_memory objects to clean up at shutdown
warnings.warn('resource_tracker: There appear to be %d '
Exception ignored in: <function GCSRecordWriter.__del__ at 0x7e67a3b05440>
Traceback (most recent call last):
File "/usr/local/lib/python3.12/site-packages/tensorboardX/record_writer.py", line 134, in __del__
File "/usr/local/lib/python3.12/site-packages/tensorboardX/record_writer.py", line 158, in close
File "/usr/local/lib/python3.12/site-packages/tensorboardX/record_writer.py", line 149, in flush
File "/usr/local/lib/python3.12/copy.py", line 87, in copy
ImportError: sys.meta_path is None, Python is likely shutting down
XPK End: Mon Apr 20 19:33:32 UTC 2026
EXIT_CODE=1