MaxView

XPK Start: Wed Apr 22 09:43:39 UTC 2026
2026-04-22 09:43:56.650817: 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)
I0422 09:44:00.198069 136752905148224 max_utils.py:273] Attempting to initialize the jax distributed system...
INFO:2026-04-22 09:44:09,239:jax._src.distributed:149: Starting JAX distributed service on [::]:8482
I0422 09:44:09.239579 136752905148224 distributed.py:149] Starting JAX distributed service on [::]:8482
INFO:2026-04-22 09:44:09,241:jax._src.distributed:166: Connecting to JAX distributed service on mt-07-distill-smoke-ionjm-slice-job-0-0.mt-07-distill-smoke-ionjm:8482
I0422 09:44:09.241882 136752905148224 distributed.py:166] Connecting to JAX distributed service on mt-07-distill-smoke-ionjm-slice-job-0-0.mt-07-distill-smoke-ionjm:8482
I0422 09:44:10.262789 136752905148224 max_utils.py:284] Jax distributed system initialized!
I0422 09:44:16.492371 136752905148224 max_utils.py:244] Jax distributed system is already initialized.
I0422 09:44:16.966160 136752905148224 max_utils.py:244] Jax distributed system is already initialized.
I0422 09:44:16.967430 136752905148224 pyconfig.py:432] Config param abort_on_inf_loss: True
I0422 09:44:16.967483 136752905148224 pyconfig.py:432] Config param abort_on_nan_loss: True
I0422 09:44:16.967511 136752905148224 pyconfig.py:432] Config param act_quantization_calibration_method: absmax
I0422 09:44:16.967535 136752905148224 pyconfig.py:432] Config param activation_dropout_for_audio: 0.0
I0422 09:44:16.967568 136752905148224 pyconfig.py:432] Config param activation_function_for_audio: gelu
I0422 09:44:16.967600 136752905148224 pyconfig.py:432] Config param activations_in_float32: False
I0422 09:44:16.967629 136752905148224 pyconfig.py:432] Config param adam_b1: 0.9
I0422 09:44:16.967660 136752905148224 pyconfig.py:432] Config param adam_b2: 0.95
I0422 09:44:16.967688 136752905148224 pyconfig.py:432] Config param adam_eps: 1e-08
I0422 09:44:16.967724 136752905148224 pyconfig.py:432] Config param adam_eps_root: 0.0
I0422 09:44:16.967752 136752905148224 pyconfig.py:432] Config param adam_weight_decay: 0.1
I0422 09:44:16.967781 136752905148224 pyconfig.py:432] Config param adamw_mask: []
I0422 09:44:16.967808 136752905148224 pyconfig.py:432] Config param add_bos: True
I0422 09:44:16.967834 136752905148224 pyconfig.py:432] Config param add_eos: True
I0422 09:44:16.967861 136752905148224 pyconfig.py:432] Config param allow_split_physical_axes: False
I0422 09:44:16.967887 136752905148224 pyconfig.py:432] Config param ar_cache_axis_order: 1,2,0,3
I0422 09:44:16.967912 136752905148224 pyconfig.py:432] Config param async_checkpointing: True
I0422 09:44:16.967939 136752905148224 pyconfig.py:432] Config param async_scheduling: False
I0422 09:44:16.967964 136752905148224 pyconfig.py:432] Config param attention: dot_product
I0422 09:44:16.967991 136752905148224 pyconfig.py:432] Config param attention_bias: False
I0422 09:44:16.968017 136752905148224 pyconfig.py:432] Config param attention_dropout_for_audio: 0.0
I0422 09:44:16.968047 136752905148224 pyconfig.py:432] Config param attention_out: RematLocation.REMAT
I0422 09:44:16.968078 136752905148224 pyconfig.py:432] Config param attention_output_dim: -1
I0422 09:44:16.968116 136752905148224 pyconfig.py:432] Config param attention_sink: False
I0422 09:44:16.968135 136752905148224 pyconfig.py:432] Config param attention_type: global
I0422 09:44:16.968151 136752905148224 pyconfig.py:432] Config param attn_logits_soft_cap: None
I0422 09:44:16.968168 136752905148224 pyconfig.py:432] Config param audio_path: 
I0422 09:44:16.968182 136752905148224 pyconfig.py:432] Config param audio_placeholder: <|audio|>
I0422 09:44:16.968198 136752905148224 pyconfig.py:432] Config param autoregressive_decode_assert: 
I0422 09:44:16.968214 136752905148224 pyconfig.py:432] Config param base_config: base.yml
I0422 09:44:16.968234 136752905148224 pyconfig.py:432] Config param base_emb_dim: 16
I0422 09:44:16.968250 136752905148224 pyconfig.py:432] Config param base_mlp_dim: 64
I0422 09:44:16.968267 136752905148224 pyconfig.py:432] Config param base_moe_mlp_dim: -1
I0422 09:44:16.968284 136752905148224 pyconfig.py:432] Config param base_num_decoder_layers: 1
I0422 09:44:16.968302 136752905148224 pyconfig.py:432] Config param base_num_kv_heads: 2
I0422 09:44:16.968318 136752905148224 pyconfig.py:432] Config param base_num_query_heads: 2
I0422 09:44:16.968333 136752905148224 pyconfig.py:432] Config param base_output_directory: 
I0422 09:44:16.968347 136752905148224 pyconfig.py:432] Config param batch_size: 1
I0422 09:44:16.968364 136752905148224 pyconfig.py:432] Config param batch_split_factor: 1
I0422 09:44:16.968381 136752905148224 pyconfig.py:432] Config param beta_fast: 32
I0422 09:44:16.968396 136752905148224 pyconfig.py:432] Config param beta_slow: 1
I0422 09:44:16.968412 136752905148224 pyconfig.py:432] Config param bwd_quantization_calibration_method: absmax
I0422 09:44:16.968433 136752905148224 pyconfig.py:432] Config param capacity_factor: -1.0
I0422 09:44:16.968461 136752905148224 pyconfig.py:432] Config param cast_logits_to_fp32: True
I0422 09:44:16.968488 136752905148224 pyconfig.py:432] Config param chat_template: 
I0422 09:44:16.968513 136752905148224 pyconfig.py:432] Config param chat_template_path: 
I0422 09:44:16.968532 136752905148224 pyconfig.py:432] Config param checkpoint_conversion_fn: None
I0422 09:44:16.968547 136752905148224 pyconfig.py:432] Config param checkpoint_dir: None
I0422 09:44:16.968565 136752905148224 pyconfig.py:432] Config param checkpoint_is_quantized: False
I0422 09:44:16.968581 136752905148224 pyconfig.py:432] Config param checkpoint_period: 2000
I0422 09:44:16.968598 136752905148224 pyconfig.py:432] Config param checkpoint_storage_concurrent_gb: 96
I0422 09:44:16.968614 136752905148224 pyconfig.py:432] Config param checkpoint_storage_target_data_file_size_bytes: 2147483648
I0422 09:44:16.968630 136752905148224 pyconfig.py:432] Config param checkpoint_storage_use_ocdbt: True
I0422 09:44:16.968646 136752905148224 pyconfig.py:432] Config param checkpoint_storage_use_zarr3: True
I0422 09:44:16.968661 136752905148224 pyconfig.py:432] Config param checkpoint_todelete_full_path: None
I0422 09:44:16.968677 136752905148224 pyconfig.py:432] Config param checkpoint_todelete_subdir: None
I0422 09:44:16.968692 136752905148224 pyconfig.py:432] Config param chips_per_vm: 4
I0422 09:44:16.968707 136752905148224 pyconfig.py:432] Config param chunk_attn_window_size: 0
I0422 09:44:16.968721 136752905148224 pyconfig.py:432] Config param collect_stack_trace: False
I0422 09:44:16.968737 136752905148224 pyconfig.py:432] Config param colocated_python_checkpointing: False
I0422 09:44:16.968751 136752905148224 pyconfig.py:432] Config param colocated_python_data_input: False
I0422 09:44:16.968766 136752905148224 pyconfig.py:432] Config param compile_topology: 
I0422 09:44:16.968780 136752905148224 pyconfig.py:432] Config param compile_topology_num_slices: -1
I0422 09:44:16.968796 136752905148224 pyconfig.py:432] Config param compile_xla_flags: 
I0422 09:44:16.968812 136752905148224 pyconfig.py:432] Config param compiled_trainstep_file: 
I0422 09:44:16.968827 136752905148224 pyconfig.py:432] Config param compute_axis_order: 0,1,2,3
I0422 09:44:16.968842 136752905148224 pyconfig.py:432] Config param constant_bound_config: []
I0422 09:44:16.968858 136752905148224 pyconfig.py:432] Config param context: RematLocation.REMAT
I0422 09:44:16.968874 136752905148224 pyconfig.py:432] Config param context_parallel_load_balance: True
I0422 09:44:16.968889 136752905148224 pyconfig.py:432] Config param context_parallel_size: 1
I0422 09:44:16.968904 136752905148224 pyconfig.py:432] Config param context_parallel_strategy: all_gather
I0422 09:44:16.968919 136752905148224 pyconfig.py:432] Config param context_sharding: context
I0422 09:44:16.968934 136752905148224 pyconfig.py:432] Config param conv_chunksize_for_audio: 500
I0422 09:44:16.968950 136752905148224 pyconfig.py:432] Config param conv_stride_for_vit: 14
I0422 09:44:16.968964 136752905148224 pyconfig.py:432] Config param cost_estimate_flops_bwd: -1
I0422 09:44:16.968980 136752905148224 pyconfig.py:432] Config param cost_estimate_flops_fwd: -1
I0422 09:44:16.968996 136752905148224 pyconfig.py:432] Config param custom_mesh: 
I0422 09:44:16.969010 136752905148224 pyconfig.py:432] Config param custom_mesh_and_rule: 
I0422 09:44:16.969026 136752905148224 pyconfig.py:432] Config param d_model_for_audio: 256
I0422 09:44:16.969039 136752905148224 pyconfig.py:432] Config param data_sharding: (('data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'context', 'context_autoregressive', 'tensor', 'tensor_transpose', 'tensor_sequence', 'expert', 'autoregressive'),)
I0422 09:44:16.969059 136752905148224 pyconfig.py:432] Config param data_shuffle_seed: 0
I0422 09:44:16.969076 136752905148224 pyconfig.py:432] Config param dataset_name: c4/en:3.0.1
I0422 09:44:16.969108 136752905148224 pyconfig.py:432] Config param dataset_path: 
I0422 09:44:16.969123 136752905148224 pyconfig.py:432] Config param dataset_type: DatasetType.HF
I0422 09:44:16.969141 136752905148224 pyconfig.py:432] Config param dcn_autoregressive_parallelism: 1
I0422 09:44:16.969155 136752905148224 pyconfig.py:432] Config param dcn_context_autoregressive_parallelism: 1
I0422 09:44:16.969170 136752905148224 pyconfig.py:432] Config param dcn_context_parallelism: 1
I0422 09:44:16.969186 136752905148224 pyconfig.py:432] Config param dcn_data_parallelism: -1
I0422 09:44:16.969203 136752905148224 pyconfig.py:432] Config param dcn_diloco_parallelism: 1
I0422 09:44:16.969218 136752905148224 pyconfig.py:432] Config param dcn_expert_parallelism: 1
I0422 09:44:16.969237 136752905148224 pyconfig.py:432] Config param dcn_fsdp_parallelism: 1
I0422 09:44:16.969253 136752905148224 pyconfig.py:432] Config param dcn_fsdp_transpose_parallelism: 1
I0422 09:44:16.969270 136752905148224 pyconfig.py:432] Config param dcn_parallelism: [1, -1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
I0422 09:44:16.969295 136752905148224 pyconfig.py:432] Config param dcn_pipeline_parallelism: 1
I0422 09:44:16.969312 136752905148224 pyconfig.py:432] Config param dcn_sequence_parallelism: 1
I0422 09:44:16.969328 136752905148224 pyconfig.py:432] Config param dcn_tensor_parallelism: 1
I0422 09:44:16.969343 136752905148224 pyconfig.py:432] Config param dcn_tensor_sequence_parallelism: 1
I0422 09:44:16.969357 136752905148224 pyconfig.py:432] Config param dcn_tensor_transpose_parallelism: 1
I0422 09:44:16.969373 136752905148224 pyconfig.py:432] Config param debug: {'rl': False}
I0422 09:44:16.969389 136752905148224 pyconfig.py:432] Config param debug_sharding: False
I0422 09:44:16.969403 136752905148224 pyconfig.py:432] Config param decode_sampling_nucleus_p: -1
I0422 09:44:16.969419 136752905148224 pyconfig.py:432] Config param decode_sampling_strategy: SamplingStrategy.GREEDY
I0422 09:44:16.969435 136752905148224 pyconfig.py:432] Config param decode_sampling_temperature: 1.0
I0422 09:44:16.969451 136752905148224 pyconfig.py:432] Config param decode_sampling_top_k: 0
I0422 09:44:16.969465 136752905148224 pyconfig.py:432] Config param decoder_block: DecoderBlockType.GPT3
I0422 09:44:16.969481 136752905148224 pyconfig.py:432] Config param decoder_layer_input: RematLocation.DEVICE
I0422 09:44:16.969496 136752905148224 pyconfig.py:432] Config param deepstack_visual_indexes_for_vit: []
I0422 09:44:16.969511 136752905148224 pyconfig.py:432] Config param degenerate_group_masking: True
I0422 09:44:16.969527 136752905148224 pyconfig.py:432] Config param dense_init_scale: 1.0
I0422 09:44:16.969542 136752905148224 pyconfig.py:432] Config param diloco_outer_lr: 0.3
I0422 09:44:16.969559 136752905148224 pyconfig.py:432] Config param diloco_outer_momentum: 0.9
I0422 09:44:16.969575 136752905148224 pyconfig.py:432] Config param diloco_sync_period: 36
I0422 09:44:16.969589 136752905148224 pyconfig.py:432] Config param distill_alpha: 0.5
I0422 09:44:16.969605 136752905148224 pyconfig.py:432] Config param distill_alpha_end: None
I0422 09:44:16.969622 136752905148224 pyconfig.py:432] Config param distill_alpha_schedule: constant
I0422 09:44:16.969637 136752905148224 pyconfig.py:432] Config param distill_beta: 0.0
I0422 09:44:16.969654 136752905148224 pyconfig.py:432] Config param distill_beta_end: None
I0422 09:44:16.969669 136752905148224 pyconfig.py:432] Config param distill_beta_schedule: constant
I0422 09:44:16.969683 136752905148224 pyconfig.py:432] Config param distill_feature_loss_type: cosine
I0422 09:44:16.969699 136752905148224 pyconfig.py:432] Config param distill_layer_indices: None
I0422 09:44:16.969713 136752905148224 pyconfig.py:432] Config param distill_temperature: 1.0
I0422 09:44:16.969728 136752905148224 pyconfig.py:432] Config param distill_temperature_end: None
I0422 09:44:16.969744 136752905148224 pyconfig.py:432] Config param distill_temperature_schedule: constant
I0422 09:44:16.969760 136752905148224 pyconfig.py:432] Config param downsample_hidden_size_for_audio: 256
I0422 09:44:16.969774 136752905148224 pyconfig.py:432] Config param dpo_beta: 0.1
I0422 09:44:16.969789 136752905148224 pyconfig.py:432] Config param dpo_label_smoothing: 0.0
I0422 09:44:16.969804 136752905148224 pyconfig.py:432] Config param dq_reduction_steps: 0
I0422 09:44:16.969819 136752905148224 pyconfig.py:432] Config param dropout_rate: 0.0
I0422 09:44:16.969833 136752905148224 pyconfig.py:432] Config param dtype: bfloat16
I0422 09:44:16.969864 136752905148224 pyconfig.py:432] Config param dtype_mm: float32
I0422 09:44:16.969880 136752905148224 pyconfig.py:432] Config param dump_hlo: False
I0422 09:44:16.969896 136752905148224 pyconfig.py:432] Config param dump_hlo_delete_local_after: True
I0422 09:44:16.969910 136752905148224 pyconfig.py:432] Config param dump_hlo_gcs_dir: gpt3-52k_2026-04-22-09-44/xla_dump
I0422 09:44:16.969924 136752905148224 pyconfig.py:432] Config param dump_hlo_local_dir: /tmp/xla_dump/
I0422 09:44:16.969940 136752905148224 pyconfig.py:432] Config param dump_hlo_local_module_name: jit_train_step
I0422 09:44:16.969954 136752905148224 pyconfig.py:432] Config param dump_hlo_module_name: jit_train_step
I0422 09:44:16.969970 136752905148224 pyconfig.py:432] Config param dump_hlo_upload_all: False
I0422 09:44:16.969984 136752905148224 pyconfig.py:432] Config param dump_hlo_xla_flags: 
I0422 09:44:16.969999 136752905148224 pyconfig.py:432] Config param dump_jaxpr: False
I0422 09:44:16.970013 136752905148224 pyconfig.py:432] Config param dump_jaxpr_delete_local_after: True
I0422 09:44:16.970029 136752905148224 pyconfig.py:432] Config param dump_jaxpr_gcs_dir: gpt3-52k_2026-04-22-09-44/jaxpr_dump
I0422 09:44:16.970045 136752905148224 pyconfig.py:432] Config param dump_jaxpr_local_dir: /tmp/jaxpr_dump/
I0422 09:44:16.970059 136752905148224 pyconfig.py:432] Config param dump_step: -1
I0422 09:44:16.970075 136752905148224 pyconfig.py:432] Config param elastic_enabled: False
I0422 09:44:16.970089 136752905148224 pyconfig.py:432] Config param elastic_max_retries: 10
I0422 09:44:16.970113 136752905148224 pyconfig.py:432] Config param elastic_timeout_seconds: 300
I0422 09:44:16.970128 136752905148224 pyconfig.py:432] Config param emb_dim: 16
I0422 09:44:16.970144 136752905148224 pyconfig.py:432] Config param enable_autocheckpoint: False
I0422 09:44:16.970157 136752905148224 pyconfig.py:432] Config param enable_checkpoint_cloud_logger: False
I0422 09:44:16.970173 136752905148224 pyconfig.py:432] Config param enable_checkpointing: True
I0422 09:44:16.970187 136752905148224 pyconfig.py:432] Config param enable_continuous_checkpointing: False
I0422 09:44:16.970202 136752905148224 pyconfig.py:432] Config param enable_data_shuffling: True
I0422 09:44:16.970216 136752905148224 pyconfig.py:432] Config param enable_diloco: False
I0422 09:44:16.970238 136752905148224 pyconfig.py:432] Config param enable_dp_attention: False
I0422 09:44:16.970253 136752905148224 pyconfig.py:432] Config param enable_dropout: False
I0422 09:44:16.970267 136752905148224 pyconfig.py:432] Config param enable_emergency_checkpoint: False
I0422 09:44:16.970282 136752905148224 pyconfig.py:432] Config param enable_expert_parallel: False
I0422 09:44:16.970296 136752905148224 pyconfig.py:432] Config param enable_gcp_goodput_metrics: True
I0422 09:44:16.970310 136752905148224 pyconfig.py:432] Config param enable_gcp_step_deviation_metrics: True
I0422 09:44:16.970325 136752905148224 pyconfig.py:432] Config param enable_goodput_recording: False
I0422 09:44:16.970340 136752905148224 pyconfig.py:432] Config param enable_jax_profiler: False
I0422 09:44:16.970356 136752905148224 pyconfig.py:432] Config param enable_llm_inference_pool: False
I0422 09:44:16.970370 136752905148224 pyconfig.py:432] Config param enable_model_warmup: False
I0422 09:44:16.970386 136752905148224 pyconfig.py:432] Config param enable_multi_tier_checkpointing: False
I0422 09:44:16.970399 136752905148224 pyconfig.py:432] Config param enable_nnx: False
I0422 09:44:16.970415 136752905148224 pyconfig.py:432] Config param enable_orbax_v1: False
I0422 09:44:16.970429 136752905148224 pyconfig.py:432] Config param enable_padding_causal_mask: True
I0422 09:44:16.970445 136752905148224 pyconfig.py:432] Config param enable_pathways_goodput: False
I0422 09:44:16.970459 136752905148224 pyconfig.py:432] Config param enable_prefix_caching: False
I0422 09:44:16.970474 136752905148224 pyconfig.py:432] Config param enable_rampup_batch_size: False
I0422 09:44:16.970488 136752905148224 pyconfig.py:432] Config param enable_single_controller: False
I0422 09:44:16.970504 136752905148224 pyconfig.py:432] Config param enable_single_replica_ckpt_restoring: False
I0422 09:44:16.970518 136752905148224 pyconfig.py:432] Config param enable_tensorboard: True
I0422 09:44:16.970533 136752905148224 pyconfig.py:432] Config param enable_tunix_perf_metrics: False
I0422 09:44:16.970547 136752905148224 pyconfig.py:432] Config param encoder_attention_heads_for_audio: 4
I0422 09:44:16.970563 136752905148224 pyconfig.py:432] Config param encoder_ffn_dim_for_audio: 512
I0422 09:44:16.970577 136752905148224 pyconfig.py:432] Config param encoder_layers_for_audio: 2
I0422 09:44:16.970592 136752905148224 pyconfig.py:432] Config param engram: RematLocation.REMAT
I0422 09:44:16.970608 136752905148224 pyconfig.py:432] Config param engram_head_dim: 1280
I0422 09:44:16.970624 136752905148224 pyconfig.py:432] Config param engram_kernel_size: 4
I0422 09:44:16.970638 136752905148224 pyconfig.py:432] Config param engram_layers: []
I0422 09:44:16.970653 136752905148224 pyconfig.py:432] Config param engram_max_ngram_size: 3
I0422 09:44:16.970667 136752905148224 pyconfig.py:432] Config param engram_num_heads: 8
I0422 09:44:16.970682 136752905148224 pyconfig.py:432] Config param engram_seed: 0
I0422 09:44:16.970697 136752905148224 pyconfig.py:432] Config param engram_vocab_bases: []
I0422 09:44:16.970712 136752905148224 pyconfig.py:432] Config param epsilon_high: None
I0422 09:44:16.970728 136752905148224 pyconfig.py:432] Config param eval_corr_lst: False
I0422 09:44:16.970744 136752905148224 pyconfig.py:432] Config param eval_data_columns: ['text']
I0422 09:44:16.970758 136752905148224 pyconfig.py:432] Config param eval_dataset_name: c4/en:3.0.1
I0422 09:44:16.970774 136752905148224 pyconfig.py:432] Config param eval_image_column: image
I0422 09:44:16.970789 136752905148224 pyconfig.py:432] Config param eval_interval: -1
I0422 09:44:16.970804 136752905148224 pyconfig.py:432] Config param eval_make_lst: False
I0422 09:44:16.970818 136752905148224 pyconfig.py:432] Config param eval_per_device_batch_size: 2
I0422 09:44:16.970834 136752905148224 pyconfig.py:432] Config param eval_sampling_strategy: greedy
I0422 09:44:16.970849 136752905148224 pyconfig.py:432] Config param eval_split: validation
I0422 09:44:16.970863 136752905148224 pyconfig.py:432] Config param eval_steps: -1
I0422 09:44:16.970879 136752905148224 pyconfig.py:432] Config param expansion_factor_real_data: -1.0
I0422 09:44:16.970895 136752905148224 pyconfig.py:432] Config param final_logits_soft_cap: None
I0422 09:44:16.970909 136752905148224 pyconfig.py:432] Config param first_num_dense_layers: 0
I0422 09:44:16.970924 136752905148224 pyconfig.py:432] Config param float32_gate_logits: False
I0422 09:44:16.970938 136752905148224 pyconfig.py:432] Config param float32_logits: False
I0422 09:44:16.970953 136752905148224 pyconfig.py:432] Config param float32_qk_product: False
I0422 09:44:16.970970 136752905148224 pyconfig.py:432] Config param float32_weight_sum: True
I0422 09:44:16.970985 136752905148224 pyconfig.py:432] Config param force_q_layout: False
I0422 09:44:16.970999 136752905148224 pyconfig.py:432] Config param force_unroll: False
I0422 09:44:16.971015 136752905148224 pyconfig.py:432] Config param freeze_audio_encoder_params: True
I0422 09:44:16.971031 136752905148224 pyconfig.py:432] Config param freeze_vision_encoder_params: True
I0422 09:44:16.971045 136752905148224 pyconfig.py:432] Config param fused_mlp: False
I0422 09:44:16.971067 136752905148224 pyconfig.py:432] Config param fused_qkv: True
I0422 09:44:16.971091 136752905148224 pyconfig.py:432] Config param gcs_metrics: False
I0422 09:44:16.971126 136752905148224 pyconfig.py:432] Config param gdn_chunk_size: 64
I0422 09:44:16.971151 136752905148224 pyconfig.py:432] Config param gdn_conv_kernel_dim: 4
I0422 09:44:16.971176 136752905148224 pyconfig.py:432] Config param gdn_key_head_dim: 128
I0422 09:44:16.971199 136752905148224 pyconfig.py:432] Config param gdn_num_key_heads: 16
I0422 09:44:16.971224 136752905148224 pyconfig.py:432] Config param gdn_num_value_heads: 32
I0422 09:44:16.971252 136752905148224 pyconfig.py:432] Config param gdn_value_head_dim: 128
I0422 09:44:16.971276 136752905148224 pyconfig.py:432] Config param generate_padding_batch_eval: False
I0422 09:44:16.971300 136752905148224 pyconfig.py:432] Config param generate_padding_batch_train: False
I0422 09:44:16.971324 136752905148224 pyconfig.py:432] Config param generate_slice: v5e-16
I0422 09:44:16.971348 136752905148224 pyconfig.py:432] Config param generation_configs: {}
I0422 09:44:16.971372 136752905148224 pyconfig.py:432] Config param global_batch_size_to_eval_on: 64
I0422 09:44:16.971397 136752905148224 pyconfig.py:432] Config param global_batch_size_to_load: 512
I0422 09:44:16.971421 136752905148224 pyconfig.py:432] Config param global_batch_size_to_load_eval: 64
I0422 09:44:16.971445 136752905148224 pyconfig.py:432] Config param global_batch_size_to_load_increment: None
I0422 09:44:16.971469 136752905148224 pyconfig.py:432] Config param global_batch_size_to_load_start: None
I0422 09:44:16.971493 136752905148224 pyconfig.py:432] Config param global_batch_size_to_train_on: 512
I0422 09:44:16.971518 136752905148224 pyconfig.py:432] Config param global_head_dim: 0
I0422 09:44:16.971541 136752905148224 pyconfig.py:432] Config param global_num_kv_heads: 0
I0422 09:44:16.971565 136752905148224 pyconfig.py:432] Config param global_parameter_scale: 1
I0422 09:44:16.971589 136752905148224 pyconfig.py:432] Config param global_rampup_samples: 500
I0422 09:44:16.971613 136752905148224 pyconfig.py:432] Config param global_rope_max_timescale: -1
I0422 09:44:16.971637 136752905148224 pyconfig.py:432] Config param global_rope_proportion: 0.25
I0422 09:44:16.971663 136752905148224 pyconfig.py:432] Config param goodput_upload_interval_seconds: 30
I0422 09:44:16.971688 136752905148224 pyconfig.py:432] Config param grad_dtype: float32
I0422 09:44:16.971740 136752905148224 pyconfig.py:432] Config param gradient_accumulation_steps: 8
I0422 09:44:16.971768 136752905148224 pyconfig.py:432] Config param gradient_clipping_threshold: 1.0
I0422 09:44:16.971794 136752905148224 pyconfig.py:432] Config param grain_data_source_max_workers: 16
I0422 09:44:16.971819 136752905148224 pyconfig.py:432] Config param grain_eval_files: 
I0422 09:44:16.971844 136752905148224 pyconfig.py:432] Config param grain_file_type: arrayrecord
I0422 09:44:16.971868 136752905148224 pyconfig.py:432] Config param grain_num_threads: 16
I0422 09:44:16.971893 136752905148224 pyconfig.py:432] Config param grain_num_threads_eval: 16
I0422 09:44:16.971918 136752905148224 pyconfig.py:432] Config param grain_packing_type: first_fit
I0422 09:44:16.971944 136752905148224 pyconfig.py:432] Config param grain_per_worker_buffer_size: 1
I0422 09:44:16.971969 136752905148224 pyconfig.py:432] Config param grain_per_worker_buffer_size_eval: 1
I0422 09:44:16.971993 136752905148224 pyconfig.py:432] Config param grain_prefetch_buffer_size: 500
I0422 09:44:16.972018 136752905148224 pyconfig.py:432] Config param grain_prefetch_buffer_size_eval: 500
I0422 09:44:16.972042 136752905148224 pyconfig.py:432] Config param grain_ram_budget_mb: 1024
I0422 09:44:16.972068 136752905148224 pyconfig.py:432] Config param grain_shuffle_buffer_size: 100
I0422 09:44:16.972105 136752905148224 pyconfig.py:432] Config param grain_train_files: 
I0422 09:44:16.972133 136752905148224 pyconfig.py:432] Config param grain_train_mixture_config_path: 
I0422 09:44:16.972158 136752905148224 pyconfig.py:432] Config param grain_worker_count: 1
I0422 09:44:16.972183 136752905148224 pyconfig.py:432] Config param grain_worker_count_eval: 1
I0422 09:44:16.972208 136752905148224 pyconfig.py:432] Config param grpo_beta: 0.08
I0422 09:44:16.972240 136752905148224 pyconfig.py:432] Config param grpo_epsilon: 0.2
I0422 09:44:16.972267 136752905148224 pyconfig.py:432] Config param hardware: tpu
I0422 09:44:16.972293 136752905148224 pyconfig.py:432] Config param hbm_utilization_vllm: 0.72
I0422 09:44:16.972319 136752905148224 pyconfig.py:432] Config param head_dim: 8
I0422 09:44:16.972343 136752905148224 pyconfig.py:432] Config param heartbeat_reporting_interval_in_seconds: 5
I0422 09:44:16.972369 136752905148224 pyconfig.py:432] Config param hf_data_dir: None
I0422 09:44:16.972394 136752905148224 pyconfig.py:432] Config param hf_eval_files: None
I0422 09:44:16.972440 136752905148224 pyconfig.py:432] Config param hf_eval_split: None
I0422 09:44:16.972462 136752905148224 pyconfig.py:432] Config param hf_name: None
I0422 09:44:16.972483 136752905148224 pyconfig.py:432] Config param hf_path: OptimalScale/ClimbMix
I0422 09:44:16.972506 136752905148224 pyconfig.py:432] Config param hf_train_files: None
I0422 09:44:16.972527 136752905148224 pyconfig.py:432] Config param hidden_size_for_vit: 1408
I0422 09:44:16.972549 136752905148224 pyconfig.py:432] Config param hide_profiler_step_metric: False
I0422 09:44:16.972570 136752905148224 pyconfig.py:432] Config param ici_autoregressive_parallelism: 1
I0422 09:44:16.972591 136752905148224 pyconfig.py:432] Config param ici_context_autoregressive_parallelism: 1
I0422 09:44:16.972613 136752905148224 pyconfig.py:432] Config param ici_context_parallelism: 1
I0422 09:44:16.972635 136752905148224 pyconfig.py:432] Config param ici_data_parallelism: 1
I0422 09:44:16.972656 136752905148224 pyconfig.py:432] Config param ici_diloco_parallelism: 1
I0422 09:44:16.972680 136752905148224 pyconfig.py:432] Config param ici_expert_parallelism: 1
I0422 09:44:16.972704 136752905148224 pyconfig.py:432] Config param ici_fsdp_parallelism: -1
I0422 09:44:16.972729 136752905148224 pyconfig.py:432] Config param ici_fsdp_transpose_parallelism: 1
I0422 09:44:16.972754 136752905148224 pyconfig.py:432] Config param ici_parallelism: [1, 1, 1, -1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
I0422 09:44:16.972780 136752905148224 pyconfig.py:432] Config param ici_pipeline_parallelism: 1
I0422 09:44:16.972805 136752905148224 pyconfig.py:432] Config param ici_sequence_parallelism: 1
I0422 09:44:16.972830 136752905148224 pyconfig.py:432] Config param ici_tensor_parallelism: 1
I0422 09:44:16.972856 136752905148224 pyconfig.py:432] Config param ici_tensor_sequence_parallelism: 1
I0422 09:44:16.972880 136752905148224 pyconfig.py:432] Config param ici_tensor_transpose_parallelism: 1
I0422 09:44:16.972905 136752905148224 pyconfig.py:432] Config param image_path: 
I0422 09:44:16.972929 136752905148224 pyconfig.py:432] Config param image_placeholder: <|image|>
I0422 09:44:16.972954 136752905148224 pyconfig.py:432] Config param image_size_for_vit: 896
I0422 09:44:16.972978 136752905148224 pyconfig.py:432] Config param indexer_head_dim: 128
I0422 09:44:16.973003 136752905148224 pyconfig.py:432] Config param indexer_loss_scaling_factor: 0.0
I0422 09:44:16.973029 136752905148224 pyconfig.py:432] Config param indexer_n_heads: 64
I0422 09:44:16.973053 136752905148224 pyconfig.py:432] Config param indexer_sparse_training: False
I0422 09:44:16.973077 136752905148224 pyconfig.py:432] Config param indexer_topk: 2048
I0422 09:44:16.973114 136752905148224 pyconfig.py:432] Config param inference_benchmark_test: False
I0422 09:44:16.973141 136752905148224 pyconfig.py:432] Config param inference_metadata_file: 
I0422 09:44:16.973167 136752905148224 pyconfig.py:432] Config param inference_microbenchmark_log_file_path: 
I0422 09:44:16.973191 136752905148224 pyconfig.py:432] Config param inference_microbenchmark_loop_iters: 10
I0422 09:44:16.973216 136752905148224 pyconfig.py:432] Config param inference_microbenchmark_num_samples: [1, 2, 3, 4, 5]
I0422 09:44:16.973250 136752905148224 pyconfig.py:432] Config param inference_microbenchmark_prefill_lengths: 64,128,256,512,1024
I0422 09:44:16.973276 136752905148224 pyconfig.py:432] Config param inference_microbenchmark_stages: prefill,generate
I0422 09:44:16.973301 136752905148224 pyconfig.py:432] Config param inference_server: MaxtextInterleavedServer
I0422 09:44:16.973326 136752905148224 pyconfig.py:432] Config param inhomogeneous_layer_cycle_interval: 1
I0422 09:44:16.973351 136752905148224 pyconfig.py:432] Config param init_weights_seed: 0
I0422 09:44:16.973376 136752905148224 pyconfig.py:432] Config param input_data_sharding_logical_axes: ['activation_embed_and_logits_batch', 'activation_norm_length']
I0422 09:44:16.973402 136752905148224 pyconfig.py:432] Config param interleave_moe_layer_step: 1
I0422 09:44:16.973427 136752905148224 pyconfig.py:432] Config param intermediate_size_for_vit: 5632
I0422 09:44:16.973453 136752905148224 pyconfig.py:432] Config param internal_compile: False
I0422 09:44:16.973477 136752905148224 pyconfig.py:432] Config param internal_compile_num_devices: -1
I0422 09:44:16.973501 136752905148224 pyconfig.py:432] Config param jax_cache_dir: ~/jax_cache
I0422 09:44:16.973518 136752905148224 pyconfig.py:432] Config param jax_debug_log_modules: 
I0422 09:44:16.973534 136752905148224 pyconfig.py:432] Config param jax_distributed_initialization_timeout: 300
I0422 09:44:16.973549 136752905148224 pyconfig.py:432] Config param jax_profiler_port: 9999
I0422 09:44:16.973564 136752905148224 pyconfig.py:432] Config param key_proj: RematLocation.REMAT
I0422 09:44:16.973581 136752905148224 pyconfig.py:432] Config param kv_cache_buffer: 256
I0422 09:44:16.973596 136752905148224 pyconfig.py:432] Config param kv_lora_rank: 512
I0422 09:44:16.973610 136752905148224 pyconfig.py:432] Config param kv_quant_axis: KvQuantAxis.HEADS_AND_DKV
I0422 09:44:16.973629 136752905148224 pyconfig.py:432] Config param kv_quant_dtype: int8
I0422 09:44:16.973643 136752905148224 pyconfig.py:432] Config param kv_wa_proj: RematLocation.REMAT
I0422 09:44:16.973659 136752905148224 pyconfig.py:432] Config param learning_rate: 0.0002
I0422 09:44:16.973681 136752905148224 pyconfig.py:432] Config param learning_rate_final_fraction: 0.1
I0422 09:44:16.973707 136752905148224 pyconfig.py:432] Config param learning_rate_schedule_steps: 200000
I0422 09:44:16.973733 136752905148224 pyconfig.py:432] Config param load_balance_loss_weight: 0.0
I0422 09:44:16.973756 136752905148224 pyconfig.py:432] Config param load_checkpoint_only_once: False
I0422 09:44:16.973778 136752905148224 pyconfig.py:432] Config param load_from_prefill_dir: False
I0422 09:44:16.973793 136752905148224 pyconfig.py:432] Config param load_full_state_path: 
I0422 09:44:16.973808 136752905148224 pyconfig.py:432] Config param load_parameters_path: gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items
I0422 09:44:16.973824 136752905148224 pyconfig.py:432] Config param local_checkpoint_directory: 
I0422 09:44:16.973838 136752905148224 pyconfig.py:432] Config param local_checkpoint_period: 0
I0422 09:44:16.973852 136752905148224 pyconfig.py:432] Config param local_rope_max_timescale: -1
I0422 09:44:16.973866 136752905148224 pyconfig.py:432] Config param local_rope_proportion: 1.0
I0422 09:44:16.973882 136752905148224 pyconfig.py:432] Config param log_config: True
I0422 09:44:16.973896 136752905148224 pyconfig.py:432] Config param log_period: 10
I0422 09:44:16.973912 136752905148224 pyconfig.py:432] Config param logical_axis_rules: (('activation_embed_and_logits_batch', ('data', 'stage', 'fsdp', 'fsdp_transpose', 'expert')), ('activation_embed_and_logits_batch_sequence', ('data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('activation_vocab', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('activation_vocab', ('tensor', 'tensor_transpose')), ('activation_vocab', 'tensor_sequence'), ('activation_vocab', ('sequence', 'context')), ('vocab', ('tensor', 'tensor_transpose', 'tensor_sequence', 'autoregressive')), ('embed_vocab', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('activation_heads', ('tensor', 'tensor_transpose', 'sequence', 'tensor_sequence', 'autoregressive')), ('activation_kv_heads', ('tensor', 'tensor_transpose', 'sequence', 'tensor_sequence')), ('activation_attn_length', ('sequence', 'context')), ('activation_attn_length', ('context',)), ('activation_q_length', ('context',)), ('activation_kv_length', ()), ('activation_attn_embed', ('tensor', 'tensor_transpose')), ('activation_kv', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('activation_kv_batch', ('data', 'fsdp', 'fsdp_transpose', 'expert')), ('activation_kv_head_dim', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('heads', ('tensor', 'tensor_transpose', 'tensor_sequence', 'autoregressive')), ('q_heads', ('tensor', 'tensor_transpose', 'tensor_sequence', 'autoregressive')), ('kv_heads', ('tensor', 'tensor_transpose', 'tensor_sequence', 'autoregressive')), ('qkv', ()), ('kv', ()), ('kv_head_dim', ()), ('q_lora', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'tensor_transpose', 'expert')), ('q_lora', ('fsdp', 'sequence', 'context', 'tensor_transpose', 'expert')), ('q_lora', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('q_lora', ('fsdp', 'sequence', 'context', 'expert')), ('q_lora_up_proj', ()), ('kv_lora', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'tensor_transpose', 'expert')), ('kv_lora', ('fsdp', 'sequence', 'context', 'tensor_transpose', 'expert')), ('kv_lora', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('kv_lora', ('fsdp', 'sequence', 'context', 'expert')), ('kv_lora_up_proj', ()), ('activation_batch_moe', ('data', 'fsdp', 'fsdp_transpose')), ('activation_length_moe', ('sequence', 'context')), ('activation_length_moe', ('context',)), ('activation_norm_length_moe', ('tensor_sequence', 'context', 'sequence')), ('activation_embed_moe', ('tensor', 'tensor_transpose')), ('activation_mlp_moe', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('activation_exp', ('expert',)), ('exp', 'expert'), ('mlp_moe', ('fsdp_transpose', 'tensor', 'tensor_sequence', 'autoregressive')), ('embed_moe', ('fsdp', 'fsdp_transpose', 'sequence', 'tensor_transpose', 'context')), ('embed_moe', ('fsdp', 'sequence', 'tensor_transpose', 'context')), ('embed_moe', ('fsdp', 'fsdp_transpose', 'sequence', 'context')), ('embed_moe', ('fsdp', 'sequence', 'context')), ('activation_mlp', ('tensor', 'tensor_transpose', 'tensor_sequence')), ('activation_batch', ('data', 'fsdp', 'fsdp_transpose', 'expert')), ('activation_length', ('sequence', 'context')), ('activation_length', ('context',)), ('activation_norm_length', ('tensor_sequence', 'context', 'sequence')), ('activation_embed', ('tensor', 'tensor_transpose')), ('activation_stage', 'stage'), ('mlp', ('fsdp_transpose', 'tensor', 'tensor_sequence', 'autoregressive')), ('embed', ('fsdp', 'fsdp_transpose', 'sequence', 'tensor_transpose', 'context', 'expert')), ('embed', ('fsdp', 'sequence', 'tensor_transpose', 'context', 'expert')), ('embed', ('fsdp', 'fsdp_transpose', 'sequence', 'context', 'expert')), ('embed', ('fsdp', 'sequence', 'context', 'expert')), ('norm', ('tensor', 'tensor_transpose')), ('layers', 'stage'), ('diloco', 'diloco'), ('engram_dim', ('tensor',)), ('dense_layers', ()), ('moe_layers', ()), ('mhc', ()), ('prefill_activation_length', ('sequence', 'context')), ('prefill_activation_norm_length', ('tensor_sequence', 'context', 'sequence')), ('activation_prefill_kv_batch', ('data', 'fsdp', 'fsdp_transpose', 'expert')), ('decode_batch', ('data', 'fsdp', 'fsdp_transpose', 'expert')), ('decode_length', ('sequence',)), ('cache_heads', ('autoregressive', 'tensor', 'tensor_transpose', 'tensor_sequence')), ('cache_heads', ('autoregressive', 'tensor', 'tensor_sequence')), ('paged_kv_heads', ('tensor',)), ('cache_batch_prefill', ()), ('cache_batch', ()), ('cache_heads_none', ()), ('cache_kv', ()), ('cache_sequence', ()), ('num_pages', ()), ('tokens_per_page', ()), ('paged_kv_head_dim_size', ()), ('mlp_no_fsdp', ('tensor', 'tensor_sequence', 'autoregressive')), ('embed_tensor_transpose', ('tensor_transpose',)), ('exp_with_fsdp', 'fsdp'))
I0422 09:44:16.974013 136752905148224 pyconfig.py:432] Config param logits_dot_in_fp32: False
I0422 09:44:16.974029 136752905148224 pyconfig.py:432] Config param logits_via_embedding: True
I0422 09:44:16.974045 136752905148224 pyconfig.py:432] Config param lora_input_adapters_path: 
I0422 09:44:16.974061 136752905148224 pyconfig.py:432] Config param loss_algo: grpo
I0422 09:44:16.974076 136752905148224 pyconfig.py:432] Config param lr_schedule_type: LearningRateScheduleType.COSINE
I0422 09:44:16.974105 136752905148224 pyconfig.py:432] Config param managed_mldiagnostics: False
I0422 09:44:16.974121 136752905148224 pyconfig.py:432] Config param managed_mldiagnostics_dir: None
I0422 09:44:16.974135 136752905148224 pyconfig.py:432] Config param managed_mldiagnostics_run_group: 
I0422 09:44:16.974151 136752905148224 pyconfig.py:432] Config param matmul_precision: MatmulPrecision.DEFAULT
I0422 09:44:16.974169 136752905148224 pyconfig.py:432] Config param max_checkify: False
I0422 09:44:16.974184 136752905148224 pyconfig.py:432] Config param max_concurrency: 256
I0422 09:44:16.974199 136752905148224 pyconfig.py:432] Config param max_corpus_chars: 10000000
I0422 09:44:16.974215 136752905148224 pyconfig.py:432] Config param max_num_batched_tokens: None
I0422 09:44:16.974235 136752905148224 pyconfig.py:432] Config param max_num_checkpoints_to_keep: None
I0422 09:44:16.974251 136752905148224 pyconfig.py:432] Config param max_num_images_per_example: -1
I0422 09:44:16.974266 136752905148224 pyconfig.py:432] Config param max_num_seqs: None
I0422 09:44:16.974282 136752905148224 pyconfig.py:432] Config param max_position_embeddings: 163840
I0422 09:44:16.974297 136752905148224 pyconfig.py:432] Config param max_prefill_predict_length: 64
I0422 09:44:16.974314 136752905148224 pyconfig.py:432] Config param max_sample_len_for_audio: 10000
I0422 09:44:16.974330 136752905148224 pyconfig.py:432] Config param max_segments_per_seq: -1
I0422 09:44:16.974346 136752905148224 pyconfig.py:432] Config param max_source_positions_for_audio: 1500
I0422 09:44:16.974361 136752905148224 pyconfig.py:432] Config param max_target_length: 2048
I0422 09:44:16.974376 136752905148224 pyconfig.py:432] Config param max_timescale_for_audio: 10000.0
I0422 09:44:16.974392 136752905148224 pyconfig.py:432] Config param megablox: True
I0422 09:44:16.974415 136752905148224 pyconfig.py:432] Config param merge_gating_gmm: False
I0422 09:44:16.974441 136752905148224 pyconfig.py:432] Config param mesh_axes: ['diloco', 'data', 'stage', 'fsdp', 'fsdp_transpose', 'sequence', 'context', 'context_autoregressive', 'tensor', 'tensor_transpose', 'tensor_sequence', 'expert', 'autoregressive']
I0422 09:44:16.974468 136752905148224 pyconfig.py:432] Config param metrics_dir: None
I0422 09:44:16.974493 136752905148224 pyconfig.py:432] Config param metrics_file: 
I0422 09:44:16.974518 136752905148224 pyconfig.py:432] Config param mhc_expansion_rate: 1
I0422 09:44:16.974542 136752905148224 pyconfig.py:432] Config param micro_batch_size_to_eval_on: 64
I0422 09:44:16.974565 136752905148224 pyconfig.py:432] Config param micro_batch_size_to_train_on: 64
I0422 09:44:16.974587 136752905148224 pyconfig.py:432] Config param mla_kv: RematLocation.REMAT
I0422 09:44:16.974611 136752905148224 pyconfig.py:432] Config param mla_naive_kvcache: True
I0422 09:44:16.974635 136752905148224 pyconfig.py:432] Config param mla_q: RematLocation.REMAT
I0422 09:44:16.974657 136752905148224 pyconfig.py:432] Config param mlp_activations: ['gelu']
I0422 09:44:16.974678 136752905148224 pyconfig.py:432] Config param mlp_activations_limit: -1.0
I0422 09:44:16.974702 136752905148224 pyconfig.py:432] Config param mlp_bias: False
I0422 09:44:16.974726 136752905148224 pyconfig.py:432] Config param mlp_dim: 64
I0422 09:44:16.974750 136752905148224 pyconfig.py:432] Config param mlpwi: RematLocation.REMAT
I0422 09:44:16.974775 136752905148224 pyconfig.py:432] Config param mlpwi_0: RematLocation.REMAT
I0422 09:44:16.974800 136752905148224 pyconfig.py:432] Config param mlpwi_1: RematLocation.REMAT
I0422 09:44:16.974825 136752905148224 pyconfig.py:432] Config param mlpwo: RematLocation.REMAT
I0422 09:44:16.974850 136752905148224 pyconfig.py:432] Config param moba: False
I0422 09:44:16.974875 136752905148224 pyconfig.py:432] Config param moba_chunk_size: 1024
I0422 09:44:16.974900 136752905148224 pyconfig.py:432] Config param moba_topk: 8
I0422 09:44:16.974924 136752905148224 pyconfig.py:432] Config param model_call_mode: 
I0422 09:44:16.974949 136752905148224 pyconfig.py:432] Config param model_name: gpt3-52k
I0422 09:44:16.974973 136752905148224 pyconfig.py:432] Config param moe_expert_input_dim: -1
I0422 09:44:16.974997 136752905148224 pyconfig.py:432] Config param moe_fsdp_use_two_stage_all_gather: False
I0422 09:44:16.975022 136752905148224 pyconfig.py:432] Config param moe_mlp_dim: -1
I0422 09:44:16.975048 136752905148224 pyconfig.py:432] Config param moe_mlpwi_0: RematLocation.REMAT
I0422 09:44:16.975065 136752905148224 pyconfig.py:432] Config param moe_mlpwi_1: RematLocation.REMAT
I0422 09:44:16.975081 136752905148224 pyconfig.py:432] Config param moe_mlpwo: RematLocation.REMAT
I0422 09:44:16.975106 136752905148224 pyconfig.py:432] Config param monitor_goodput: False
I0422 09:44:16.975122 136752905148224 pyconfig.py:432] Config param monitor_step_time_deviation: True
I0422 09:44:16.975136 136752905148224 pyconfig.py:432] Config param mrope_section: [24, 20, 20]
I0422 09:44:16.975152 136752905148224 pyconfig.py:432] Config param mscale: 1.0
I0422 09:44:16.975169 136752905148224 pyconfig.py:432] Config param mtc_data_parallelism: 0
I0422 09:44:16.975183 136752905148224 pyconfig.py:432] Config param mtp_eval_target_module: 0
I0422 09:44:16.975198 136752905148224 pyconfig.py:432] Config param mtp_loss_scaling_factor: 0.1
I0422 09:44:16.975215 136752905148224 pyconfig.py:432] Config param mtp_num_layers: 0
I0422 09:44:16.975234 136752905148224 pyconfig.py:432] Config param mu_dtype: float32
I0422 09:44:16.975261 136752905148224 pyconfig.py:432] Config param multi_sampling: False
I0422 09:44:16.975278 136752905148224 pyconfig.py:432] Config param multi_tier_checkpointing_backup_interval_minutes: 0
I0422 09:44:16.975294 136752905148224 pyconfig.py:432] Config param muon_beta: 0.95
I0422 09:44:16.975311 136752905148224 pyconfig.py:432] Config param muon_consistent_rms: None
I0422 09:44:16.975325 136752905148224 pyconfig.py:432] Config param muon_weight_decay: 0.0
I0422 09:44:16.975340 136752905148224 pyconfig.py:432] Config param n_routing_groups: -1
I0422 09:44:16.975356 136752905148224 pyconfig.py:432] Config param n_window_for_audio: 50
I0422 09:44:16.975370 136752905148224 pyconfig.py:432] Config param n_window_infer_for_audio: 800
I0422 09:44:16.975386 136752905148224 pyconfig.py:432] Config param nope_layer_interval: -1
I0422 09:44:16.975401 136752905148224 pyconfig.py:432] Config param norm_topk_prob: False
I0422 09:44:16.975415 136752905148224 pyconfig.py:432] Config param normalization_layer_epsilon: 1e-05
I0422 09:44:16.975439 136752905148224 pyconfig.py:432] Config param normalize_embedding_logits: False
I0422 09:44:16.975465 136752905148224 pyconfig.py:432] Config param num_attention_heads_for_vit: 16
I0422 09:44:16.975491 136752905148224 pyconfig.py:432] Config param num_batches: 4
I0422 09:44:16.975517 136752905148224 pyconfig.py:432] Config param num_channels_for_vit: 3
I0422 09:44:16.975542 136752905148224 pyconfig.py:432] Config param num_conv_layers_for_audio: 3
I0422 09:44:16.975566 136752905148224 pyconfig.py:432] Config param num_decoder_layers: 1
I0422 09:44:16.975590 136752905148224 pyconfig.py:432] Config param num_diloco_replicas: 1
I0422 09:44:16.975614 136752905148224 pyconfig.py:432] Config param num_epoch: 1
I0422 09:44:16.975639 136752905148224 pyconfig.py:432] Config param num_eval_passes: 1
I0422 09:44:16.975665 136752905148224 pyconfig.py:432] Config param num_experts: 1
I0422 09:44:16.975691 136752905148224 pyconfig.py:432] Config param num_experts_per_tok: 1
I0422 09:44:16.975706 136752905148224 pyconfig.py:432] Config param num_generations: 2
I0422 09:44:16.975722 136752905148224 pyconfig.py:432] Config param num_hidden_layers_for_vit: 34
I0422 09:44:16.975744 136752905148224 pyconfig.py:432] Config param num_iterations: 1
I0422 09:44:16.975771 136752905148224 pyconfig.py:432] Config param num_kv_heads: 2
I0422 09:44:16.975795 136752905148224 pyconfig.py:432] Config param num_layers_per_pipeline_stage: 1
I0422 09:44:16.975811 136752905148224 pyconfig.py:432] Config param num_mel_bins_for_audio: 128
I0422 09:44:16.975825 136752905148224 pyconfig.py:432] Config param num_pipeline_microbatches: -1
I0422 09:44:16.975839 136752905148224 pyconfig.py:432] Config param num_pipeline_repeats: -1
I0422 09:44:16.975855 136752905148224 pyconfig.py:432] Config param num_position_embeddings_for_vit: 1024
I0422 09:44:16.975869 136752905148224 pyconfig.py:432] Config param num_query_heads: 2
I0422 09:44:16.975885 136752905148224 pyconfig.py:432] Config param num_samplers_slices: -1
I0422 09:44:16.975899 136752905148224 pyconfig.py:432] Config param num_slices: 1
I0422 09:44:16.975914 136752905148224 pyconfig.py:432] Config param num_target_devices: 32
I0422 09:44:16.975930 136752905148224 pyconfig.py:432] Config param num_test_batches: 5
I0422 09:44:16.975944 136752905148224 pyconfig.py:432] Config param num_trainer_slices: -1
I0422 09:44:16.975960 136752905148224 pyconfig.py:432] Config param num_vocab_tiling: 1
I0422 09:44:16.975976 136752905148224 pyconfig.py:432] Config param off_policy_steps: 0
I0422 09:44:16.975991 136752905148224 pyconfig.py:432] Config param offline_data_dir: None
I0422 09:44:16.976006 136752905148224 pyconfig.py:432] Config param opt_type: OptimizerType.ADAM_PAX
I0422 09:44:16.976025 136752905148224 pyconfig.py:432] Config param optimize_mesh_for_tpu_v6e: False
I0422 09:44:16.976039 136752905148224 pyconfig.py:432] Config param optimizer_memory_host_offload: False
I0422 09:44:16.976055 136752905148224 pyconfig.py:432] Config param original_max_position_embeddings: 4096
I0422 09:44:16.976070 136752905148224 pyconfig.py:432] Config param out_hidden_size_for_vit: 512
I0422 09:44:16.976084 136752905148224 pyconfig.py:432] Config param out_proj: RematLocation.REMAT
I0422 09:44:16.976119 136752905148224 pyconfig.py:432] Config param output_dim_for_audio: 512
I0422 09:44:16.976142 136752905148224 pyconfig.py:432] Config param override_logical_axis_rules: False
I0422 09:44:16.976167 136752905148224 pyconfig.py:432] Config param override_model_config: True
I0422 09:44:16.976193 136752905148224 pyconfig.py:432] Config param packing: True
I0422 09:44:16.976212 136752905148224 pyconfig.py:432] Config param pagedattn_head_dim_alignment: 128
I0422 09:44:16.976231 136752905148224 pyconfig.py:432] Config param pagedattn_max_pages_per_group: -1
I0422 09:44:16.976246 136752905148224 pyconfig.py:432] Config param pagedattn_num_pages: 64
I0422 09:44:16.976260 136752905148224 pyconfig.py:432] Config param pagedattn_pages_per_compute_block: 4
I0422 09:44:16.976276 136752905148224 pyconfig.py:432] Config param pagedattn_tokens_per_page: 32
I0422 09:44:16.976292 136752905148224 pyconfig.py:432] Config param param_scan_axis: 1
I0422 09:44:16.976308 136752905148224 pyconfig.py:432] Config param parameter_memory_host_offload: False
I0422 09:44:16.976322 136752905148224 pyconfig.py:432] Config param partial_rotary_factor: 1.0
I0422 09:44:16.976338 136752905148224 pyconfig.py:432] Config param patch_size_for_vit: 14
I0422 09:44:16.976353 136752905148224 pyconfig.py:432] Config param penalty_incorrect_answer: -1.0
I0422 09:44:16.976367 136752905148224 pyconfig.py:432] Config param penalty_incorrect_format: -0.5
I0422 09:44:16.976383 136752905148224 pyconfig.py:432] Config param per_device_batch_size: 2
I0422 09:44:16.976398 136752905148224 pyconfig.py:432] Config param per_device_batch_size_increment: 2.0
I0422 09:44:16.976422 136752905148224 pyconfig.py:432] Config param per_device_batch_size_start: 4.0
I0422 09:44:16.976437 136752905148224 pyconfig.py:432] Config param pipeline_delay_activation_forwarding: False
I0422 09:44:16.976451 136752905148224 pyconfig.py:432] Config param pipeline_fsdp_ag_once: False
I0422 09:44:16.976466 136752905148224 pyconfig.py:432] Config param pipeline_fsdp_ag_per_repeat: False
I0422 09:44:16.976483 136752905148224 pyconfig.py:432] Config param pipeline_parallel_layers: 1
I0422 09:44:16.976497 136752905148224 pyconfig.py:432] Config param pixel_shuffle_ratio_for_vit: 0.5
I0422 09:44:16.976513 136752905148224 pyconfig.py:432] Config param posemb_type_for_vit: learn
I0422 09:44:16.976529 136752905148224 pyconfig.py:432] Config param position_id_per_seconds: 25
I0422 09:44:16.976543 136752905148224 pyconfig.py:432] Config param prefill_cache_axis_order: 1,2,0,3
I0422 09:44:16.976557 136752905148224 pyconfig.py:432] Config param prefill_cache_dir: 
I0422 09:44:16.976571 136752905148224 pyconfig.py:432] Config param prefill_chunk_size: 256
I0422 09:44:16.976586 136752905148224 pyconfig.py:432] Config param prefill_slice: v5e-16
I0422 09:44:16.976602 136752905148224 pyconfig.py:432] Config param prefix_caching_dram_byte: 100000000000
I0422 09:44:16.976616 136752905148224 pyconfig.py:432] Config param prefix_caching_hbm_byte: 10000000000
I0422 09:44:16.976631 136752905148224 pyconfig.py:432] Config param profile_cleanly: True
I0422 09:44:16.976645 136752905148224 pyconfig.py:432] Config param profile_periodically_period: -1
I0422 09:44:16.976661 136752905148224 pyconfig.py:432] Config param profile_power_events: False
I0422 09:44:16.976675 136752905148224 pyconfig.py:432] Config param profiler: ProfilerType.NONE
I0422 09:44:16.976692 136752905148224 pyconfig.py:432] Config param profiler_steps: 5
I0422 09:44:16.976707 136752905148224 pyconfig.py:432] Config param projector_dropout_for_vit: 0.0
I0422 09:44:16.976722 136752905148224 pyconfig.py:432] Config param projector_input_dim_for_vit: 4096
I0422 09:44:16.976737 136752905148224 pyconfig.py:432] Config param projector_output_dim_for_vit: 4096
I0422 09:44:16.976751 136752905148224 pyconfig.py:432] Config param prometheus_port: 0
I0422 09:44:16.976767 136752905148224 pyconfig.py:432] Config param prompt: I love to
I0422 09:44:16.976781 136752905148224 pyconfig.py:432] Config param pure_nnx: False
I0422 09:44:16.976796 136752905148224 pyconfig.py:432] Config param pure_nnx_decoder: False
I0422 09:44:16.976812 136752905148224 pyconfig.py:432] Config param q_lora_rank: 0
I0422 09:44:16.976831 136752905148224 pyconfig.py:432] Config param qk_clip_threshold: 100.0
I0422 09:44:16.976857 136752905148224 pyconfig.py:432] Config param qk_nope_head_dim: 128
I0422 09:44:16.976881 136752905148224 pyconfig.py:432] Config param qk_norm_with_scale: True
I0422 09:44:16.976905 136752905148224 pyconfig.py:432] Config param qk_rope_head_dim: 64
I0422 09:44:16.976928 136752905148224 pyconfig.py:432] Config param qkv_proj: RematLocation.REMAT
I0422 09:44:16.976953 136752905148224 pyconfig.py:432] Config param quant_cfg_path: 
I0422 09:44:16.976977 136752905148224 pyconfig.py:432] Config param quantization: QuantizationType.NONE
I0422 09:44:16.977005 136752905148224 pyconfig.py:432] Config param quantization_local_shard_count: 4
I0422 09:44:16.977029 136752905148224 pyconfig.py:432] Config param quantize_kvcache: False
I0422 09:44:16.977055 136752905148224 pyconfig.py:432] Config param query_proj: RematLocation.REMAT
I0422 09:44:16.977080 136752905148224 pyconfig.py:432] Config param query_wa_proj: RematLocation.REMAT
I0422 09:44:16.977121 136752905148224 pyconfig.py:432] Config param ragged_block_size: 256
I0422 09:44:16.977149 136752905148224 pyconfig.py:432] Config param ragged_buffer_factor: -1.0
I0422 09:44:16.977174 136752905148224 pyconfig.py:432] Config param rampup_end_step: 0
I0422 09:44:16.977199 136752905148224 pyconfig.py:432] Config param rampup_samples_per_increment_to_load: None
I0422 09:44:16.977225 136752905148224 pyconfig.py:432] Config param reasoning_end_token: </reasoning>
I0422 09:44:16.977256 136752905148224 pyconfig.py:432] Config param reasoning_start_token: <reasoning>
I0422 09:44:16.977284 136752905148224 pyconfig.py:432] Config param record_internal_nn_metrics: 0
I0422 09:44:16.977310 136752905148224 pyconfig.py:432] Config param remat_policy: full
I0422 09:44:16.977335 136752905148224 pyconfig.py:432] Config param remat_policy_for_vit: minimal
I0422 09:44:16.977360 136752905148224 pyconfig.py:432] Config param remove_size_one_mesh_axis_from_type: True
I0422 09:44:16.977385 136752905148224 pyconfig.py:432] Config param replicate_quant_scale: False
I0422 09:44:16.977410 136752905148224 pyconfig.py:432] Config param replicator_backup_interval_minutes: 0
I0422 09:44:16.977435 136752905148224 pyconfig.py:432] Config param report_heartbeat_metric_for_gcp_monitoring: False
I0422 09:44:16.977459 136752905148224 pyconfig.py:432] Config param report_performance_metric_for_gcp_monitoring: False
I0422 09:44:16.977484 136752905148224 pyconfig.py:432] Config param reshape_q: False
I0422 09:44:16.977509 136752905148224 pyconfig.py:432] Config param return_log_prob: False
I0422 09:44:16.977534 136752905148224 pyconfig.py:432] Config param reuse_example_batch: 0
I0422 09:44:16.977559 136752905148224 pyconfig.py:432] Config param reward_exact_answer: 5.0
I0422 09:44:16.977584 136752905148224 pyconfig.py:432] Config param reward_exact_format_match: 3.0
I0422 09:44:16.977608 136752905148224 pyconfig.py:432] Config param reward_partial_format_match: 0.5
I0422 09:44:16.977635 136752905148224 pyconfig.py:432] Config param reward_ratio_guess_to_answer_high: 0.5
I0422 09:44:16.977660 136752905148224 pyconfig.py:432] Config param reward_ratio_guess_to_answer_low: 0.25
I0422 09:44:16.977685 136752905148224 pyconfig.py:432] Config param reward_white_space_format_match: 1.5
I0422 09:44:16.977711 136752905148224 pyconfig.py:432] Config param rl: {'num_generations': 2, 'num_iterations': 1, 'grpo_beta': 0.08, 'grpo_epsilon': 0.2, 'loss_algo': 'grpo', 'use_agentic_rollout': False, 'max_concurrency': 256, 'off_policy_steps': 0, 'system_prompt': '', 'degenerate_group_masking': True, 'epsilon_high': None}
I0422 09:44:16.977744 136752905148224 pyconfig.py:432] Config param rollout_data_parallelism: -1
I0422 09:44:16.977769 136752905148224 pyconfig.py:432] Config param rollout_expert_parallelism: 1
I0422 09:44:16.977794 136752905148224 pyconfig.py:432] Config param rollout_micro_batch_size: -1
I0422 09:44:16.977819 136752905148224 pyconfig.py:432] Config param rollout_tensor_parallelism: -1
I0422 09:44:16.977843 136752905148224 pyconfig.py:432] Config param rope_attention_scaling: False
I0422 09:44:16.977868 136752905148224 pyconfig.py:432] Config param rope_factor: 40
I0422 09:44:16.977893 136752905148224 pyconfig.py:432] Config param rope_interleave: True
I0422 09:44:16.977918 136752905148224 pyconfig.py:432] Config param rope_linear_scaling_factor: 1.0
I0422 09:44:16.977943 136752905148224 pyconfig.py:432] Config param rope_max_timescale: 10000
I0422 09:44:16.977968 136752905148224 pyconfig.py:432] Config param rope_min_timescale: 1
I0422 09:44:16.977993 136752905148224 pyconfig.py:432] Config param rope_theta_for_vit: 10000
I0422 09:44:16.978018 136752905148224 pyconfig.py:432] Config param rope_truncate: True
I0422 09:44:16.978042 136752905148224 pyconfig.py:432] Config param rope_type: RopeType.DEFAULT
I0422 09:44:16.978071 136752905148224 pyconfig.py:432] Config param rope_use_scale: True
I0422 09:44:16.978112 136752905148224 pyconfig.py:432] Config param routed_bias: False
I0422 09:44:16.978141 136752905148224 pyconfig.py:432] Config param routed_bias_update_rate: 0.0
I0422 09:44:16.978167 136752905148224 pyconfig.py:432] Config param routed_scaling_factor: 1.0
I0422 09:44:16.978192 136752905148224 pyconfig.py:432] Config param routed_score_func: 
I0422 09:44:16.978216 136752905148224 pyconfig.py:432] Config param run_name: gpt3-52k_2026-04-22-09-44
I0422 09:44:16.978247 136752905148224 pyconfig.py:432] Config param sa_block_kv: 512
I0422 09:44:16.978272 136752905148224 pyconfig.py:432] Config param sa_block_kv_compute: 512
I0422 09:44:16.978296 136752905148224 pyconfig.py:432] Config param sa_block_kv_dkv: 512
I0422 09:44:16.978321 136752905148224 pyconfig.py:432] Config param sa_block_kv_dkv_compute: 512
I0422 09:44:16.978344 136752905148224 pyconfig.py:432] Config param sa_block_kv_dq: 512
I0422 09:44:16.978369 136752905148224 pyconfig.py:432] Config param sa_block_q: 512
I0422 09:44:16.978394 136752905148224 pyconfig.py:432] Config param sa_block_q_dkv: 512
I0422 09:44:16.978418 136752905148224 pyconfig.py:432] Config param sa_block_q_dq: 512
I0422 09:44:16.978443 136752905148224 pyconfig.py:432] Config param sa_k_layout: HEAD_DIM_MINOR
I0422 09:44:16.978466 136752905148224 pyconfig.py:432] Config param sa_q_layout: HEAD_DIM_MINOR
I0422 09:44:16.978491 136752905148224 pyconfig.py:432] Config param sa_use_fused_bwd_kernel: False
I0422 09:44:16.978516 136752905148224 pyconfig.py:432] Config param sa_v_layout: HEAD_DIM_MINOR
I0422 09:44:16.978540 136752905148224 pyconfig.py:432] Config param sampler_devices_fraction: 0.5
I0422 09:44:16.978566 136752905148224 pyconfig.py:432] Config param save_checkpoint_on_completion: True
I0422 09:44:16.978591 136752905148224 pyconfig.py:432] Config param save_config_to_gcs: False
I0422 09:44:16.978616 136752905148224 pyconfig.py:432] Config param save_quantized_params_path: 
I0422 09:44:16.978641 136752905148224 pyconfig.py:432] Config param scale_embedding_for_audio: True
I0422 09:44:16.978665 136752905148224 pyconfig.py:432] Config param scan_layers: True
I0422 09:44:16.978689 136752905148224 pyconfig.py:432] Config param scan_layers_per_stage: False
I0422 09:44:16.978713 136752905148224 pyconfig.py:432] Config param scan_pipeline_iterations: True
I0422 09:44:16.978737 136752905148224 pyconfig.py:432] Config param scan_pipeline_repeats: False
I0422 09:44:16.978761 136752905148224 pyconfig.py:432] Config param set_remat_policy_on_layers_per_stage: False
I0422 09:44:16.978786 136752905148224 pyconfig.py:432] Config param set_remat_policy_on_pipeline_iterations: True
I0422 09:44:16.978811 136752905148224 pyconfig.py:432] Config param sft_train_on_completion_only: False
I0422 09:44:16.978836 136752905148224 pyconfig.py:432] Config param shard_exp_on_fsdp: False
I0422 09:44:16.978860 136752905148224 pyconfig.py:432] Config param shard_mode: ShardMode.AUTO
I0422 09:44:16.978888 136752905148224 pyconfig.py:432] Config param shard_optimizer_over_data: False
I0422 09:44:16.978912 136752905148224 pyconfig.py:432] Config param sharding_strategy: None
I0422 09:44:16.978936 136752905148224 pyconfig.py:432] Config param sharding_tolerance: 0.02
I0422 09:44:16.978962 136752905148224 pyconfig.py:432] Config param shardy: True
I0422 09:44:16.978987 136752905148224 pyconfig.py:432] Config param share_kv_projections: False
I0422 09:44:16.979011 136752905148224 pyconfig.py:432] Config param shared_experts: 0
I0422 09:44:16.979036 136752905148224 pyconfig.py:432] Config param sinkhorn_iterations: 20
I0422 09:44:16.979060 136752905148224 pyconfig.py:432] Config param skip_first_n_steps_for_profiler: 1
I0422 09:44:16.979084 136752905148224 pyconfig.py:432] Config param skip_jax_distributed_system: False
I0422 09:44:16.979123 136752905148224 pyconfig.py:432] Config param skip_step_interval: 128
I0422 09:44:16.979148 136752905148224 pyconfig.py:432] Config param skip_step_on_spikes: False
I0422 09:44:16.979173 136752905148224 pyconfig.py:432] Config param skip_step_scaling_factor: 6.0
I0422 09:44:16.979198 136752905148224 pyconfig.py:432] Config param sliding_window_size: 0
I0422 09:44:16.979222 136752905148224 pyconfig.py:432] Config param solution_end_token: </answer>
I0422 09:44:16.979252 136752905148224 pyconfig.py:432] Config param solution_start_token: <answer>
I0422 09:44:16.979277 136752905148224 pyconfig.py:432] Config param source_checkpoint_layout: orbax
I0422 09:44:16.979302 136752905148224 pyconfig.py:432] Config param sparse_matmul: True
I0422 09:44:16.979326 136752905148224 pyconfig.py:432] Config param spatial_merge_size_for_vit: 2
I0422 09:44:16.979351 136752905148224 pyconfig.py:432] Config param stack_prefill_result_cache: False
I0422 09:44:16.979376 136752905148224 pyconfig.py:432] Config param stack_trace_interval_seconds: 600
I0422 09:44:16.979401 136752905148224 pyconfig.py:432] Config param stack_trace_to_cloud: False
I0422 09:44:16.979426 136752905148224 pyconfig.py:432] Config param step_deviation_interval_seconds: 30
I0422 09:44:16.979450 136752905148224 pyconfig.py:432] Config param steps: 200000
I0422 09:44:16.979474 136752905148224 pyconfig.py:432] Config param stop_strings: None
I0422 09:44:16.979498 136752905148224 pyconfig.py:432] Config param student_overrides: {'model_name': 'llama3.1-8b'}
I0422 09:44:16.979524 136752905148224 pyconfig.py:432] Config param student_params_to_update: None
I0422 09:44:16.979551 136752905148224 pyconfig.py:432] Config param subslice_shape: 
I0422 09:44:16.979576 136752905148224 pyconfig.py:432] Config param swap_space_vllm_gb: 2
I0422 09:44:16.979600 136752905148224 pyconfig.py:432] Config param system_prompt: 
I0422 09:44:16.979618 136752905148224 pyconfig.py:432] Config param target_eval_loss: 0.0
I0422 09:44:16.979642 136752905148224 pyconfig.py:432] Config param teacher_overrides: {'model_name': 'llama3.1-8b'}
I0422 09:44:16.979669 136752905148224 pyconfig.py:432] Config param temperature_tuning: False
I0422 09:44:16.979694 136752905148224 pyconfig.py:432] Config param temporal_patch_size_for_vit: 2
I0422 09:44:16.979719 136752905148224 pyconfig.py:432] Config param tensorboard_dir: None
I0422 09:44:16.979743 136752905148224 pyconfig.py:432] Config param tensors_on_device: None
I0422 09:44:16.979768 136752905148224 pyconfig.py:432] Config param tensors_to_offload: None
I0422 09:44:16.979792 136752905148224 pyconfig.py:432] Config param test_batch_start_index: 0
I0422 09:44:16.979817 136752905148224 pyconfig.py:432] Config param tile_size_for_vit: 336
I0422 09:44:16.979842 136752905148224 pyconfig.py:432] Config param tokenize_eval_data: True
I0422 09:44:16.979867 136752905148224 pyconfig.py:432] Config param tokenize_train_data: True
I0422 09:44:16.979891 136752905148224 pyconfig.py:432] Config param tokenizer_path: meta-llama/Llama-3.1-8B
I0422 09:44:16.979915 136752905148224 pyconfig.py:432] Config param tokenizer_type: TokenizerType.HUGGINGFACE
I0422 09:44:16.979942 136752905148224 pyconfig.py:432] Config param topk_routing_group: -1
I0422 09:44:16.979968 136752905148224 pyconfig.py:432] Config param train_data_columns: ['text']
I0422 09:44:16.979992 136752905148224 pyconfig.py:432] Config param train_fraction: 1.0
I0422 09:44:16.980017 136752905148224 pyconfig.py:432] Config param train_image_column: image
I0422 09:44:16.980042 136752905148224 pyconfig.py:432] Config param train_micro_batch_size: -1
I0422 09:44:16.980067 136752905148224 pyconfig.py:432] Config param train_split: train
I0422 09:44:16.980091 136752905148224 pyconfig.py:432] Config param trainable_parameters_mask: []
I0422 09:44:16.980129 136752905148224 pyconfig.py:432] Config param trainable_position_size: 2048
I0422 09:44:16.980155 136752905148224 pyconfig.py:432] Config param trainer_devices_fraction: 0.5
I0422 09:44:16.980181 136752905148224 pyconfig.py:432] Config param upload_all_profiler_results: False
I0422 09:44:16.980206 136752905148224 pyconfig.py:432] Config param use_2d_fsdp_sharding: False
I0422 09:44:16.980235 136752905148224 pyconfig.py:432] Config param use_agentic_rollout: False
I0422 09:44:16.980260 136752905148224 pyconfig.py:432] Config param use_audio: False
I0422 09:44:16.980286 136752905148224 pyconfig.py:432] Config param use_audio_in_video: False
I0422 09:44:16.980311 136752905148224 pyconfig.py:432] Config param use_batch_split_schedule: False
I0422 09:44:16.980336 136752905148224 pyconfig.py:432] Config param use_chat_template: False
I0422 09:44:16.980361 136752905148224 pyconfig.py:432] Config param use_chunked_prefill: False
I0422 09:44:16.980386 136752905148224 pyconfig.py:432] Config param use_custom_sort_vjp: True
I0422 09:44:16.980411 136752905148224 pyconfig.py:432] Config param use_dpo: False
I0422 09:44:16.980436 136752905148224 pyconfig.py:432] Config param use_gather_mosaic_kernel: False
I0422 09:44:16.980460 136752905148224 pyconfig.py:432] Config param use_grpo: True
I0422 09:44:16.980484 136752905148224 pyconfig.py:432] Config param use_indexer: False
I0422 09:44:16.980509 136752905148224 pyconfig.py:432] Config param use_iota_embed: True
I0422 09:44:16.980533 136752905148224 pyconfig.py:432] Config param use_jax_splash: False
I0422 09:44:16.980558 136752905148224 pyconfig.py:432] Config param use_max_logit_estimate: -1
I0422 09:44:16.980581 136752905148224 pyconfig.py:432] Config param use_mrope: False
I0422 09:44:16.980605 136752905148224 pyconfig.py:432] Config param use_multimodal: False
I0422 09:44:16.980629 136752905148224 pyconfig.py:432] Config param use_pathways: True
I0422 09:44:16.980654 136752905148224 pyconfig.py:432] Config param use_post_attn_norm: False
I0422 09:44:16.980678 136752905148224 pyconfig.py:432] Config param use_post_ffw_norm: False
I0422 09:44:16.980702 136752905148224 pyconfig.py:432] Config param use_qk_clip: False
I0422 09:44:16.980727 136752905148224 pyconfig.py:432] Config param use_qk_norm: False
I0422 09:44:16.980752 136752905148224 pyconfig.py:432] Config param use_qk_norm_in_gdn: True
I0422 09:44:16.980776 136752905148224 pyconfig.py:432] Config param use_qwix_quantization: False
I0422 09:44:16.980801 136752905148224 pyconfig.py:432] Config param use_ragged_attention: False
I0422 09:44:16.980825 136752905148224 pyconfig.py:432] Config param use_random_routing: False
I0422 09:44:16.980850 136752905148224 pyconfig.py:432] Config param use_replicator_service: False
I0422 09:44:16.980874 136752905148224 pyconfig.py:432] Config param use_ring_of_experts: False
I0422 09:44:16.980899 136752905148224 pyconfig.py:432] Config param use_sft: False
I0422 09:44:16.980924 136752905148224 pyconfig.py:432] Config param use_splash_scheduler: False
I0422 09:44:16.980948 136752905148224 pyconfig.py:432] Config param use_tokamax_gmm: False
I0422 09:44:16.980972 136752905148224 pyconfig.py:432] Config param use_tokamax_splash: False
I0422 09:44:16.980997 136752905148224 pyconfig.py:432] Config param use_truncation: True
I0422 09:44:16.981021 136752905148224 pyconfig.py:432] Config param use_tunix_gradient_accumulation: False
I0422 09:44:16.981045 136752905148224 pyconfig.py:432] Config param use_untrainable_positional_embedding: False
I0422 09:44:16.981070 136752905148224 pyconfig.py:432] Config param use_vertex_tensorboard: False
I0422 09:44:16.981105 136752905148224 pyconfig.py:432] Config param using_pipeline_parallelism: False
I0422 09:44:16.981132 136752905148224 pyconfig.py:432] Config param v_head_dim: 128
I0422 09:44:16.981159 136752905148224 pyconfig.py:432] Config param v_norm_with_scale: True
I0422 09:44:16.981185 136752905148224 pyconfig.py:432] Config param value_proj: RematLocation.REMAT
I0422 09:44:16.981212 136752905148224 pyconfig.py:432] Config param vertex_tensorboard_project: 
I0422 09:44:16.981240 136752905148224 pyconfig.py:432] Config param vertex_tensorboard_region: 
I0422 09:44:16.981265 136752905148224 pyconfig.py:432] Config param video_path: 
I0422 09:44:16.981290 136752905148224 pyconfig.py:432] Config param video_placeholder: <|video|>
I0422 09:44:16.981314 136752905148224 pyconfig.py:432] Config param vision_output_dim_for_vit: 4096
I0422 09:44:16.981338 136752905148224 pyconfig.py:432] Config param vision_output_length: -1
I0422 09:44:16.981363 136752905148224 pyconfig.py:432] Config param vllm_additional_config: {}
I0422 09:44:16.981388 136752905148224 pyconfig.py:432] Config param vllm_hf_config_path: 
I0422 09:44:16.981413 136752905148224 pyconfig.py:432] Config param vllm_hf_overrides: {}
I0422 09:44:16.981437 136752905148224 pyconfig.py:432] Config param vocab_size: 32000
I0422 09:44:16.981462 136752905148224 pyconfig.py:432] Config param warmup_steps_fraction: 0.1
I0422 09:44:16.981488 136752905148224 pyconfig.py:432] Config param weight_dtype: float32
I0422 09:44:16.981530 136752905148224 pyconfig.py:432] Config param weight_quantization_calibration_method: absmax
I0422 09:44:16.981556 136752905148224 pyconfig.py:432] Config param wi_tile_dlhs_batch_seq: 512
I0422 09:44:16.981581 136752905148224 pyconfig.py:432] Config param wi_tile_dlhs_embed_dim: 1024
I0422 09:44:16.981606 136752905148224 pyconfig.py:432] Config param wi_tile_dlhs_mlp_dim: 1024
I0422 09:44:16.981630 136752905148224 pyconfig.py:432] Config param wi_tile_drhs_batch_seq: 512
I0422 09:44:16.981654 136752905148224 pyconfig.py:432] Config param wi_tile_drhs_embed_dim: 1024
I0422 09:44:16.981679 136752905148224 pyconfig.py:432] Config param wi_tile_drhs_mlp_dim: 1024
I0422 09:44:16.981704 136752905148224 pyconfig.py:432] Config param wi_tile_fwd_batch_seq: 512
I0422 09:44:16.981728 136752905148224 pyconfig.py:432] Config param wi_tile_fwd_embed_dim: 1024
I0422 09:44:16.981753 136752905148224 pyconfig.py:432] Config param wi_tile_fwd_mlp_dim: 1024
I0422 09:44:16.981777 136752905148224 pyconfig.py:432] Config param wo_tile_dlhs_batch_seq: 512
I0422 09:44:16.981804 136752905148224 pyconfig.py:432] Config param wo_tile_dlhs_embed_dim: 1024
I0422 09:44:16.981830 136752905148224 pyconfig.py:432] Config param wo_tile_dlhs_mlp_dim: 1024
I0422 09:44:16.981854 136752905148224 pyconfig.py:432] Config param wo_tile_drhs_batch_seq: 512
I0422 09:44:16.981880 136752905148224 pyconfig.py:432] Config param wo_tile_drhs_embed_dim: 1024
I0422 09:44:16.981905 136752905148224 pyconfig.py:432] Config param wo_tile_drhs_mlp_dim: 1024
I0422 09:44:16.981930 136752905148224 pyconfig.py:432] Config param wo_tile_fwd_batch_seq: 512
I0422 09:44:16.981954 136752905148224 pyconfig.py:432] Config param wo_tile_fwd_embed_dim: 1024
I0422 09:44:16.981979 136752905148224 pyconfig.py:432] Config param wo_tile_fwd_mlp_dim: 1024
I0422 09:44:16.982003 136752905148224 pyconfig.py:432] Config param wsd_decay_steps_fraction: 0.1
I0422 09:44:16.982030 136752905148224 pyconfig.py:432] Config param wsd_decay_style: WsdDecayStyle.LINEAR
I0422 09:44:16.982058 136752905148224 pyconfig.py:432] Config param xprof_e2e_enable_fw_power_level_event: False
I0422 09:44:16.982083 136752905148224 pyconfig.py:432] Config param xprof_e2e_enable_fw_thermal_event: False
I0422 09:44:16.982121 136752905148224 pyconfig.py:432] Config param xprof_e2e_enable_fw_throttle_event: False
I0422 09:44:16.982147 136752905148224 pyconfig.py:432] Config param xprof_tpu_power_trace_level: 0
I0422 09:44:16.982175 136752905148224 pyconfig.py:432] Config param z_loss_multiplier: 0.0
I0422 09:44:16.982660 136752905148224 tokenizer.py:245] Tokenizer path: meta-llama/Llama-2-7b-chat-hf
I0422 09:44:16.982703 136752905148224 tokenizer.py:224] Loading HF tokenizer: meta-llama/Llama-2-7b-chat-hf
I0422 09:44:20.842545 136752905148224 _schedule.py:129] A polynomial schedule was set with a non-positive `transition_steps` value; this results in a constant schedule with value `init_value`.
I0422 09:44:20.845645 136752905148224 maxtext_utils.py:1718] Num_devices: 32, shape (1, 4, 1, 8, 1, 1, 1, 1, 1, 1, 1, 1, 1)
I0422 09:44:20.845774 136752905148224 train_distill.py:596] Applying logical axis rules for model initialization and training...
I0422 09:44:20.845874 136752905148224 train_distill.py:600] Loading Student from ...
I0422 09:44:20.845924 136752905148224 train_distill.py:169] --- Student Configuration ---
I0422 09:44:20.845959 136752905148224 train_distill.py:170]   Model Name:      gpt3-52k
I0422 09:44:20.845983 136752905148224 train_distill.py:171]   Dimensions:      1 Layers, 16 Emb Dim, 8 Head Dim
I0422 09:44:20.846002 136752905148224 train_distill.py:174]   Attention Heads: 2 Query, 2 KV
I0422 09:44:20.846029 136752905148224 train_distill.py:175]   Vocab Size:      32000
I0422 09:44:20.846057 136752905148224 train_distill.py:176]   Checkpoint:      
I0422 09:44:20.846086 136752905148224 train_distill.py:465] Initializing model: gpt3-52k...
I0422 09:44:22.245740 136752905148224 train_distill.py:614] Loading Teacher from gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items...
I0422 09:44:22.245848 136752905148224 train_distill.py:169] --- Teacher Configuration ---
I0422 09:44:22.245877 136752905148224 train_distill.py:170]   Model Name:      gpt3-52k
I0422 09:44:22.245910 136752905148224 train_distill.py:171]   Dimensions:      1 Layers, 16 Emb Dim, 8 Head Dim
I0422 09:44:22.245941 136752905148224 train_distill.py:174]   Attention Heads: 2 Query, 2 KV
I0422 09:44:22.245962 136752905148224 train_distill.py:175]   Vocab Size:      32000
I0422 09:44:22.245986 136752905148224 train_distill.py:176]   Checkpoint:      gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items
I0422 09:44:22.246015 136752905148224 train_distill.py:465] Initializing model: gpt3-52k...
I0422 09:44:23.317214 136752905148224 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0422 09:44:23.317648 136752905148224 base_pytree_checkpoint_handler.py:411] 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 0x7c5f8cb581d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0422 09:44:23.317708 136752905148224 abstract_checkpointer.py:35] orbax-checkpoint version: 0.11.28
W0422 09:44:23.828176 136752905148224 checkpoint.py:202] Metadata file does not exist: gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items/_CHECKPOINT_METADATA
I0422 09:44:24.374187    2140 google_auth_provider.cc:181] Running on GCE, using service account 562977990677-compute@developer.gserviceaccount.com
I0422 09:44:25.491559 136752905148224 checkpointer.py:304] Restoring checkpoint from gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items.
W0422 09:44:27.656772 136752905148224 transform_utils.py:230] The transformations API will eventually be replaced by an upgraded design. The current API will not be removed until this point, but it will no longer be actively worked on.
I0422 09:44:27.657161 136752905148224 transform_utils.py:288] The following keys are not loaded from the original tree after applying specified transforms: params/params/decoder/to_nnx__rngs/aqt/count, params/params/decoder/to_nnx__rngs/aqt/key, params/params/decoder/to_nnx__rngs/dropout/count, params/params/decoder/to_nnx__rngs/dropout/key, params/params/decoder/to_nnx__rngs/params/count, params/params/decoder/to_nnx__rngs/params/key
I0422 09:44:28.474496 136752905148224 checkpointer.py:318] Finished restoring checkpoint in 3.35 seconds from gs://lance-maxtext/pt_seed_ckpts/pt_seed_ckpts/pt_seed_ckpt_gpt352k_v32k_linen/checkpoints/4/items.
I0422 09:44:29.151560 136752905148224 train_distill.py:640] Initializing Data Iterators via MaxText pipeline...
I0422 09:44:29.215008 136752905148224 config.py:112] TensorFlow version 2.20.0 available.
I0422 09:44:29.215517 136752905148224 config.py:125] JAX version 0.8.3 available.
E0422 09:44:31.256362 136752905148224 packing.py:209] PackAndBatchOperation is deprecated. Please use lazy_dataset.FirstFitPackIterDataset instead.
I0422 09:44:31.256577 136752905148224 data_loader.py:408] Adding CopyNumPyArrayToSharedMemory MapTransform.
I0422 09:44:31.259731 136752905148224 train_distill.py:410] Input Pipeline Checkpointing: DISABLED
I0422 09:44:31.259792 136752905148224 train_distill.py:414] Reason: Iterator 'MultiHostDataLoadIterator' is not recognized as Grain (dataset_type='DatasetType.HF', has_save=False)
I0422 09:44:31.259856 136752905148224 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0422 09:44:31.259935 136752905148224 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 0x7c5f8cb581d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0422 09:44:31.259977 136752905148224 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0422 09:44:31.260007 136752905148224 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 0x7c5f8cb581d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0422 09:44:31.260048 136752905148224 checkpoint_manager.py:702] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=None, item_handlers={'model_params': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6343320>, 'optimizer_state': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6343290>, 'custom_metadata': <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a6343200>}, handler_registry=None
I0422 09:44:31.260255 136752905148224 composite_checkpoint_handler.py:237] Deferred registration for item: "model_params". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6343320>` 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`.
I0422 09:44:31.260299 136752905148224 composite_checkpoint_handler.py:237] Deferred registration for item: "optimizer_state". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6343290>` 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`.
I0422 09:44:31.260328 136752905148224 composite_checkpoint_handler.py:237] Deferred registration for item: "custom_metadata". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a6343200>` 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`.
I0422 09:44:31.260358 136752905148224 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c5588da2f30>` 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`.
I0422 09:44:31.260386 136752905148224 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 0x7c48a6343320>, ('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6343320>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6343290>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6343290>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a6343200>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a6343200>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c5588da2f30>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c5588da2f30>}).
I0422 09:44:31.260788 136752905148224 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c48a61e4040> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0422 09:44:33.630902 136752905148224 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260422_093114/pt_distill_linen_xpk_feat_nnx_trainstate_and_training_loop_20260422_093114_07_distill_smoke/checkpoints
I0422 09:44:33.942277 136752905148224 checkpoint_manager.py:921] [process=6][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=2000, 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_trainstate_and_training_loop_20260422_093114/pt_distill_linen_xpk_feat_nnx_trainstate_and_training_loop_20260422_093114_07_distill_smoke/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c48a63431d0>
I0422 09:44:33.942458 136752905148224 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0422 09:44:33.942531 136752905148224 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 0x7c5f8cb581d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0422 09:44:33.942568 136752905148224 pytree_checkpoint_handler.py:577] save_device_host_concurrent_bytes=None
I0422 09:44:33.942600 136752905148224 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 0x7c5f8cb581d0>, enable_pinned_host_transfer=False, save_concurrent_bytes: 96000000000 (89.4 GiB), restore_concurrent_bytes: 96000000000 (89.4 GiB)
I0422 09:44:33.942635 136752905148224 checkpoint_manager.py:1983] [process=6][thread=MainThread][wait_until_finished] No Save Finalize thread to wait for. Returning.
I0422 09:44:33.942690 136752905148224 checkpoint.py:459] Closing _NonBlockingMetadataStore(enable_write=True, _write_lock=<locked _thread.RLock object owner=136752905148224 count=1 at 0x7c48a6ad3800>, _store_impl=<orbax.checkpoint._src.metadata.checkpoint._MetadataStoreImpl object at 0x7c48a6342fc0>, _single_thread_executor=<concurrent.futures.thread.ThreadPoolExecutor object at 0x7c48a6342f90>, _write_futures=[])
I0422 09:44:33.943039 136752905148224 checkpoint.py:459] Closing _NonBlockingMetadataStore(enable_write=True, _write_lock=<locked _thread.RLock object owner=136752905148224 count=1 at 0x7c48a6ad3800>, _store_impl=<orbax.checkpoint._src.metadata.checkpoint._MetadataStoreImpl object at 0x7c48a6342fc0>, _single_thread_executor=<concurrent.futures.thread.ThreadPoolExecutor object at 0x7c48a6342f90>, _write_futures=[])
I0422 09:44:33.943066 136752905148224 checkpoint.py:459] Closing _NonBlockingMetadataStore(enable_write=True, _write_lock=<locked _thread.RLock object owner=136752905148224 count=1 at 0x7c48a6ad3800>, _store_impl=<orbax.checkpoint._src.metadata.checkpoint._MetadataStoreImpl object at 0x7c48a6342fc0>, _single_thread_executor=<concurrent.futures.thread.ThreadPoolExecutor object at 0x7c48a6342f90>, _write_futures=[])
I0422 09:44:33.943110 136752905148224 checkpoint_manager.py:702] [process=6][thread=MainThread] CheckpointManager init: checkpointers=None, item_names=None, item_handlers={'model_params': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a63431a0>, 'optimizer_state': <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6342600>, 'custom_metadata': <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a5f65d30>, 'iter': <maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7c48a5f66510>}, handler_registry=None
I0422 09:44:33.943218 136752905148224 composite_checkpoint_handler.py:237] Deferred registration for item: "model_params". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a63431a0>` 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`.
I0422 09:44:33.943251 136752905148224 composite_checkpoint_handler.py:237] Deferred registration for item: "optimizer_state". Adding handler `<orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6342600>` 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`.
I0422 09:44:33.943275 136752905148224 composite_checkpoint_handler.py:237] Deferred registration for item: "custom_metadata". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a5f65d30>` 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`.
I0422 09:44:33.943303 136752905148224 composite_checkpoint_handler.py:237] Deferred registration for item: "iter". Adding handler `<maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7c48a5f66510>` for item "iter" and save args `<class 'maxtext.common.checkpointing.GrainCheckpointSave'>` and restore args `<class 'maxtext.common.checkpointing.GrainCheckpointRestore'>` to `_handler_registry`.
I0422 09:44:33.943327 136752905148224 composite_checkpoint_handler.py:237] Deferred registration for item: "metrics". Adding handler `<orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a5f66750>` 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`.
I0422 09:44:33.943358 136752905148224 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 0x7c48a63431a0>, ('model_params', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a63431a0>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeSaveArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6342600>, ('optimizer_state', <class 'orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeRestoreArgs'>): <orbax.checkpoint._src.handlers.pytree_checkpoint_handler.PyTreeCheckpointHandler object at 0x7c48a6342600>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a5f65d30>, ('custom_metadata', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a5f65d30>, ('iter', <class 'maxtext.common.checkpointing.GrainCheckpointSave'>): <maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7c48a5f66510>, ('iter', <class 'maxtext.common.checkpointing.GrainCheckpointRestore'>): <maxtext.common.checkpointing.GrainCheckpointHandler object at 0x7c48a5f66510>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonSaveArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a5f66750>, ('metrics', <class 'orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonRestoreArgs'>): <orbax.checkpoint._src.handlers.json_checkpoint_handler.JsonCheckpointHandler object at 0x7c48a5f66750>}).
I0422 09:44:33.943427 136752905148224 async_checkpointer.py:177] [process=6][thread=MainThread] Using barrier_sync_fn: <function get_barrier_sync_fn.<locals>._fn at 0x7c48a61e4220> timeout: 600 secs and primary_host=0 for async checkpoint writes
I0422 09:44:34.751224 136752905148224 checkpoint_manager.py:1788] Found 0 checkpoint steps in gs://lance-maxtext/pt_ckpt_xpk_feat_nnx_trainstate_and_training_loop_20260422_093114/pt_distill_linen_xpk_feat_nnx_trainstate_and_training_loop_20260422_093114_07_distill_smoke/checkpoints
I0422 09:44:34.759547 136752905148224 checkpoint_manager.py:921] [process=6][thread=MainThread] CheckpointManager created,  primary_host=0, CheckpointManagerOptions=CheckpointManagerOptions(save_interval_steps=2000, 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_trainstate_and_training_loop_20260422_093114/pt_distill_linen_xpk_feat_nnx_trainstate_and_training_loop_20260422_093114_07_distill_smoke/checkpoints: <orbax.checkpoint.checkpoint_manager.CheckpointManager object at 0x7c48a6342840>
I0422 09:44:34.759980 136752905148224 train_distill.py:691] Starting Distillation Training...
I0422 09:44:34.760082 136752905148224 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))
I0422 09:44:35.490928 136752905148224 peft_trainer.py:600] Compiled train_step cache size: 0

Training:   0%|          | 0/5 [00:00<?, ?step/s]I0422 09:44:35.492703 136608716142336 grain_pool.py:367] Grain pool will use 1 processes.
I0422 09:44:35.519465 136608716142336 grain_pool.py:440] Grain pool will start child processes.
I0422 09:44:35.524662 136608716142336 grain_pool.py:448] Grain pool started all child processes.
2026-04-22 09:44:41.585758: 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)
I0422 09:44:45.073842 136752905148224 utils.py:86] Train loop finished in: 9.5823 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/distillation/train_distill.py", line 765, 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/distillation/train_distill.py", line 761, in main
    train_distill(student_config, teacher_config, is_offline, global_config.offline_data_dir)
  File "/deps/src/maxtext/trainers/post_train/distillation/train_distill.py", line 693, in train_distill
    trainer.train(train_iter, eval_iter)
  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=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=9, process_index=2, coords=(1,2,0), core_on_chip=0), TpuDevice(id=2, process_index=1, coords=(2,0,0), 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=22, process_index=5, coords=(2,5,0), core_on_chip=0), TpuDevice(id=30, process_index=7, coords=(2,7,0), core_on_chip=0), TpuDevice(id=6, process_index=1, coords=(2,1,0), core_on_chip=0), TpuDevice(id=31, process_index=7, coords=(3,7,0), 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=23, process_index=5, coords=(3,5,0), core_on_chip=0), TpuDevice(id=12, process_index=2, coords=(0,3,0), core_on_chip=0), TpuDevice(id=11, process_index=3, coords=(3,2,0), core_on_chip=0), TpuDevice(id=13, process_index=2, coords=(1,3,0), 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=1, process_index=0, coords=(1,0,0), core_on_chip=0), TpuDevice(id=18, process_index=5, coords=(2,4,0), core_on_chip=0), TpuDevice(id=14, process_index=3, coords=(2,3,0), core_on_chip=0), TpuDevice(id=26, process_index=7, coords=(2,6,0), core_on_chip=0), TpuDevice(id=3, process_index=1, coords=(3,0,0), core_on_chip=0), TpuDevice(id=7, process_index=1, coords=(3,1,0), core_on_chip=0), TpuDevice(id=5, process_index=0, coords=(1,1,0), core_on_chip=0), TpuDevice(id=8, process_index=2, coords=(0,2,0), core_on_chip=0), TpuDevice(id=4, process_index=0, coords=(0,1,0), 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)}
I0422 09:44:45.422266 136608716142336 grain_pool.py:542] Grain pool is exiting.
I0422 09:44:45.422365 136608716142336 grain_pool.py:547] Shutting down multiprocessing system.
I0422 09:44:46.869148 136608716142336 grain_pool.py:547] Shutting down multiprocessing system.

Training:   0%|          | 0/5 [00:13<?, ?step/s]
/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 '
XPK End: Wed Apr 22 09:44:54 UTC 2026
EXIT_CODE=1