rm.yaml

# Bradley-Terry (BT) Reward Model Training Configuration
rm:
  ## total number of steps to train will equal
  ## min((max_num_epochs * len(train_dataloader)), max_num_steps)
  max_num_epochs: 1
  max_num_steps: -1  # by default, train for 1 epoch

  val_period: 16
  val_batches: -1
  val_global_batch_size: 32
  val_micro_batch_size: 1
  val_at_start: false
  val_at_end: false
  seed: 42

checkpointing:
  enabled: true
  checkpoint_dir: "results/rm"
  metric_name: "val:validation-default_loss" # one of "val:" or "train:" followed by the metric name
  higher_is_better: false
  keep_top_k: 3
  save_period: ${rm.val_period}
  checkpoint_must_save_by: null

policy:
  model_name: "meta-llama/Llama-3.2-1B-Instruct"
  tokenizer:
    name: ${policy.model_name} ## specify if you'd like to use a tokenizer different from the model's default
    # We don't use the "default" chat template because the Llama tokenizer inserts the current
    # date in the system prompt, which could make the reward model's output date-dependent.
    chat_template: "{{- bos_token }}\n\n{#- This block extracts the system message, so we can slot it into the right place. #}\n{%- if messages[0]['role'] == 'system' %}\n    {%- set system_message = messages[0]['content']|trim %}\n    {%- set messages = messages[1:] %}\n{%- else %}\n    {%- set system_message = '' %}\n{%- endif %}\n\n{#- System message #}\n{{- '<|start_header_id|>system<|end_header_id|>\n\n' }}\n{{- system_message }}\n{{- '<|eot_id|>' }}\n\n{%- for message in messages %}\n    {{- '<|start_header_id|>' + message['role'] + '<|end_header_id|>\n\n' + message['content'] | trim + '<|eot_id|>' }}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|start_header_id|>assistant<|end_header_id>\n\n' }}\n{%- endif %}"
    chat_template_kwargs: null # can be used to pass kwargs to the chat template, e.g., enable_thinking=true
  train_global_batch_size: 128
  train_micro_batch_size: 1
  max_total_sequence_length: 8192
  precision: "bfloat16"
  activation_checkpointing_enabled: false

  offload_optimizer_for_logprob: false

  reward_model_cfg:
    enabled: true  # loads model as a Reward Model (do not change)
    reward_model_type: "bradley_terry"  # only "bradley_terry" is currently supported

  dtensor_cfg:
    enabled: true
    cpu_offload: false
    sequence_parallel: false
    activation_checkpointing: false
    tensor_parallel_size: 1
    context_parallel_size: 1
    custom_parallel_plan: null

  dynamic_batching:
    enabled: false

  sequence_packing:
    enabled: false

  # makes the training sequence length divisible by the tensor parallel size
  # this is useful for sequence parallel training
  make_sequence_length_divisible_by: ${policy.dtensor_cfg.tensor_parallel_size}
  max_grad_norm: 1.0

  optimizer:
    name: "torch.optim.AdamW"
    kwargs:
      lr: 2.0e-6
      weight_decay: 0.1
      betas: [0.9, 0.98]
      eps: 1e-5
      # when using Dtensor, we need to set `foreach` and `fused` to false
      foreach: false
      fused: false
    
  ## ignored since enabled=false, but needed for testing purposes
  megatron_cfg:
    enabled: false
    empty_unused_memory_level: 1
    activation_checkpointing: false
    tensor_model_parallel_size: 2
    pipeline_model_parallel_size: 2
    context_parallel_size: 1
    pipeline_dtype: ${policy.precision}
    num_layers_in_first_pipeline_stage: null
    num_layers_in_last_pipeline_stage: null
    sequence_parallel: false

    optimizer:
      optimizer: "adam"
      lr: 2.0e-6
      min_lr: 1.9999e-6
      weight_decay: 0.1
      bf16: false
      fp16: false
      params_dtype: "float32"

      #adam
      adam_beta1: 0.9
      adam_beta2: 0.98
      adam_eps: 1e-5

      #sgd
      sgd_momentum: 0.9

      #distributed optimizer
      use_distributed_optimizer: true
      use_precision_aware_optimizer: true

      clip_grad: ${policy.max_grad_norm}

      # optimizer cpu offload
      optimizer_cpu_offload: false
      optimizer_offload_fraction: 0.0

    scheduler:
      start_weight_decay: ${policy.megatron_cfg.optimizer.weight_decay}
      end_weight_decay: ${policy.megatron_cfg.optimizer.weight_decay}
      weight_decay_incr_style: "constant"
      lr_decay_style: "constant"
      lr_decay_iters: 1000
      lr_warmup_iters: 50
      lr_warmup_init: 1.9999e-6

    distributed_data_parallel_config:
      grad_reduce_in_fp32: false
      overlap_grad_reduce: true
      overlap_param_gather: false
      data_parallel_sharding_strategy: "optim_grads_params"

    fp8_cfg:
      enabled: false
      fp8: "e4m3"
      fp8_recipe: "blockwise"
      fp8_param: false
    
data:
  max_input_seq_length: ${policy.max_total_sequence_length}
  shuffle: true
  num_workers: 1

  # dataset
  train:
    dataset_name: HelpSteer3
    split: train
  validation:
    dataset_name: HelpSteer3
    split: validation
  # You can use custom preference datasets for training and validation. For example:
  # 1. PreferenceDataset
  #   train:
  #     # this dataset will override prompt_key and use the default values for other vars
  #     data_path: /path/to/local/train_dataset.jsonl  # local file or hf_org/hf_dataset_name (HuggingFace)
  #     subset: null  # used for HuggingFace datasets
  #     split: train  # used for HuggingFace datasets
  #   validation:
  #     # this dataset will use the default values for other vars except data_path
  #     data_path: /path/to/local/val_dataset.jsonl
  #   default:
  #     # will use below vars as default values if dataset doesn't specify it
  #     dataset_name: PreferenceDataset
  #     prompt_file: null
  #     system_prompt_file: null
  #   # multiple validation sets is supported by using val_data_paths
  #   # this will be removed after refactor
  #   val_data_paths:
  #     <NameOfValidationDataset1>: /path/to/local/val_dataset_1.jsonl
  #     <NameOfValidationDataset2>: /path/to/local/val_dataset_2.jsonl
  # 2. BinaryPreferenceDataset
  #   train:
  #     # this dataset will override prompt_key and use the default values for other vars
  #     data_path: /path/to/local/train_dataset.jsonl  # local file or hf_org/hf_dataset_name (HuggingFace)
  #     prompt_key: context
  #     subset: null  # used for HuggingFace datasets
  #     split: train  # used for HuggingFace datasets
  #   validation:
  #     # this dataset will use the default values for other vars except data_path
  #     data_path: /path/to/local/val_dataset.jsonl
  #   default:
  #     # will use below vars as default values if dataset doesn't specify it
  #     dataset_name: BinaryPreferenceDataset
  #     prompt_key: prompt
  #     chosen_key: chosen
  #     rejected_key: rejected
  #     prompt_file: null
  #     system_prompt_file: null
  # See https://github.com/NVIDIA-NeMo/RL/blob/main/docs/guides/rm.md#datasets for more details.

  # If you are doing checkpointing, `metric_name` should reflect the metric and validation set to be tracked. For example:
  #   checkpointing:
  #     metric_name: "validation-<NameOfValidationDataset1>_loss"
  #   ...

logger:
  log_dir: "logs"  # Base directory for all logs
  wandb_enabled: true # Make sure you do a ``wandb login [Your API key]'' before running
  tensorboard_enabled: true
  mlflow_enabled: false
  swanlab_enabled: false # Disable SwanLab logging
  monitor_gpus: true  # If true, will monitor GPU usage and log to wandb and/or tensorboard
  wandb:
    project: "rm-dev"
    name: "rm-dev-${data.train.dataset_name}"
  swanlab:
    project: "rm-dev"
    name: "rm-dev-${data.train.dataset_name}"
  tensorboard:
    log_dir: "tb_logs-rm-dev-${data.train.dataset_name}"
  gpu_monitoring:
    collection_interval: 10  # How often to collect GPU usage metrics (in seconds)
    flush_interval: 10  # How often to flush GPU usage metrics to the loggers (in seconds)

cluster:
  gpus_per_node: 1
  num_nodes: 1