cant save checkpoint

[Heiheicxx]

W0117 14:48:50.972433 140051385186112 experiment.py:220] Your experiment's self.CHECKPOINT_ATTRS and self.NON_BROADCAST_CHECKPOINT_ATTRS are empty. Your job will not checkpoint any state or parameters.
I0117 14:48:50.972585 140051385186112 utils.py:615] Saved checkpoint latest with id 0.
I0117 14:48:50.973322 140051385186112 utils.py:316] [jaxline] training loop finished.
I0117 14:48:50.974981 140051385186112 utils.py:309] [jaxline] final checkpoint starting...
W0117 14:48:50.975140 140051385186112 experiment.py:220] Your experiment's self.CHECKPOINT_ATTRS and self.NON_BROADCAST_CHECKPOINT_ATTRS are empty. Your job will not checkpoint any state or parameters.
I0117 14:48:50.975250 140051385186112 utils.py:615] Saved checkpoint latest with id 1.
I0117 14:48:50.975420 140051385186112 utils.py:316] [jaxline] final checkpoint finished.
I0117 14:48:50.975657 140051385186112 utils.py:498] Waiting for a periodic action to finish...
I0117 14:49:06.433700 140051385186112 utils.py:309] [jaxline] rendezvous starting...
I0117 14:49:06.486311 140051385186112 utils.py:316] [jaxline] rendezvous finished.

train config as follow:
I0117 14:44:59.102810 140051385186112 train.py:76] Training with config:
best_checkpoint_all_hosts: false
best_model_eval_metric: ''
best_model_eval_metric_higher_is_better: true
binary_args:

• !!python/tuple
  • --define cudnn_embed_so
  • 1
• !!python/tuple
  • --define=cuda_compress
  • 1
    checkpoint_dir: /workplace/0_code/2_Game/2025/ai-imgbit/training/
    checkpoint_interval_type: null
    eval_initial_weights: false
    eval_modes: !!python/tuple []
    eval_specific_checkpoint_dir: /workplace/0_code/2_Game/2025/ai-imgbit/eval/
    experiment_kwargs:
    config:
    dataset:
    name: clic2020
    num_examples: 1
    num_frames: null
    root_dir: /workplace/0_code/2_Game/2025/ai-imgbit/clic2020
    skip_examples: 0
    spatial_patch_size: null
    video_idx: null
    eval: {}
    log_gradient_norms: false
    log_per_datum_metrics: true
    loss:
    rd_weight: 0.001
    rd_weight_warmup_steps: 0
    model:
    entropy:
    activation_fn: gelu
    clip_like_cool_chic: true
    conditional_spec:
    interpolation: bilinear
    prev_kernel_shape: &id001 !!python/tuple
    - 3
    - 3
    use_conditioning: true
    use_prev_grid: true
    context_num_rows_cols: *id001
    layers: &id002 !!python/tuple
    - 24
    - 24
    scale_range: !!python/tuple
    - 0.001
    - 150
    shift_log_scale: 8.0
    use_linear_w_init: true
    latents:
    add_gains: true
    downsampling_exponents: !!python/tuple
    - 0
    - 1
    - 2
    - 3
    - 4
    - 5
    - 6
    downsampling_factor: 2.0
    gain_factor: null
    gain_values: null
    learnable_gains: false
    num_grids: 7
    q_step: 0.4
    quant:
    q_steps_bias:
    - 5.0e-05
    - 0.0001
    - 0.0005
    - 0.001
    - 0.003
    - 0.006
    - 0.01
    q_steps_weight:
    - 5.0e-05
    - 0.0001
    - 0.0005
    - 0.001
    - 0.003
    - 0.006
    - 0.01
    synthesis:
    activation_fn: gelu
    add_activation_before_residual: false
    add_layer_norm: false
    b_last_init_input_mean: false
    clip_range: !!python/tuple
    - 0.0
    - 1.0
    kernel_shape: 1
    layers: *id002
    num_residual_layers: 2
    residual_kernel_shape: 3
    upsampling:
    kwargs:
    interpolation_method: bilinear
    type: image_resize
    opt:
    cosine_decay_schedule: true
    cosine_decay_schedule_kwargs:
    alpha: 0.0
    decay_steps: 100
    init_value: 0.01
    grad_norm_clip: 0.1
    max_num_ste_steps: 100
    noise_log_every: 100
    num_noise_steps: 100
    ste_break_at_lr: 1.0e-08
    ste_init_lr: 0.0001
    ste_log_every: 50
    ste_lr_decay_factor: 0.8
    ste_num_steps_not_improved: 20
    ste_reset_params_at_lr_decay: true
    ste_uses_cosine_decay: false
    quant:
    kumaraswamy_decay_steps: 100
    kumaraswamy_end_value: 1.0
    kumaraswamy_init_value: 2.0
    noise_quant_type: soft_round
    soft_round_temp_end: 0.1
    soft_round_temp_start: 0.3
    ste_quant_type: ste_soft_round
    ste_soft_round_temp: 0.0001
    use_kumaraswamy_noise: true
    interval_type: steps
    legacy_random_seed_behavior: false
    log_all_hosts: false
    log_all_train_data: false
    log_async: true
    log_tensors_interval: 1
    log_train_data_interval: 1.0
    logging_interval_type: null
    max_checkpoints_to_keep: 5
    one_off_evaluate: false
    periodic_action_growth_ratios: null
    random_mode_eval: same_host_same_device
    random_mode_train: unique_host_unique_device
    random_seed: 0
    save_checkpoint_interval: 1
    save_initial_train_checkpoint: false
    train_checkpoint_all_hosts: false
    training_steps: 1