Direct Preference Optimization (DPO)¶

DPO (Direct Preference Optimization) is an elegant simplification of RLHF (Reinforcement Learning from Human Feedback) that makes preference learning more computationally efficient, especially for large language models.

The two key innovations are:

Eliminating the reward model: Instead of training a separate reward model to score outputs (which requires additional compute and memory), DPO directly optimizes the policy using preference data. It reparameterizes the reward function implicitly through the policy itself, deriving a closed-form solution for the optimal policy.
Preference-based optimization: DPO treats the preference learning problem as a classification task over pairs of responses. It maximizes the likelihood that preferred responses are ranked higher than rejected ones under the current policy, relative to a reference policy. This approach eliminates the need for sampling and reward model queries during training.

These changes are particularly valuable for LLM training because they reduce computational overhead by removing the need for a separate reward model and RL training loop, provide more stable training dynamics by avoiding the complexities of reinforcement learning, and they simplify implementation while achieving comparable or better performance than traditional RLHF.

Example¶

from agilerl.algorithms import DPO
from agilerl.llm_envs import PreferenceGym
from accelerate import Accelerator
from datasets import load_dataset
from peft import get_peft_model
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Instantiate the model and the associated tokenizer
model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen2.5-3B",
    torch_dtype=torch.bfloat16,
)
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-3B")

# Instantiate an accelerator object for distributed training
accelerator = Accelerator()

# Load the dataset into a PreferenceGym environment
raw_dataset = load_dataset("HumanLLMs/Human-Like-DPO-Dataset", split="train").shuffle(seed=42)
train_test_split = raw_dataset.train_test_split(test_size=0.1)
train_dataset = train_test_split["train"]
test_dataset = train_test_split["test"]
env = PreferenceGym(
  train_dataset=train_dataset,
  test_dataset=test_dataset,
  tokenizer=tokenizer,
  data_batch_size_per_gpu=16,
  accelerator=accelerator,
)

# Instantiate the agent
agent = DPO(
  env.observation_space,
  env.action_space,
  actor_network=model,
  pad_token_id=tokenizer.eos_token_id,
  pad_token=tokenizer.eos_token,
  device="cuda" if torch.cuda.is_available() else "cpu",
  batch_size=32,
  lr=0.000005,
  beta=0.001,
  update_epochs=1,
  seed=42,
  accelerator=accelerator,
)

Training a DPO agent¶

To train a DPO agent on a single preference gym environment, use the finetune_llm_preference function:

from agilerl.training.train_llm import finetune_llm_preference

finetune_llm_preference(
  [agent],
  env,
  num_epochs=1,
  checkpoint_steps=250,
  accelerator=accelerator,
)

Saving and Loading Agents¶

To save an agent, use the save_llm_checkpoint function:

from agilerl.utils.utils import save_llm_checkpoint

save_llm_checkpoint(agent, "path/to/checkpoint")

To load a trained model, you must use the HuggingFace .from_pretrained method, AgileRL is compatible with HuggingFace and Peft models:

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

base_model = AutoModelForCausalLM.from_pretrained(
    "Qwen/Qwen2.5-3B",
    torch_dtype=torch.bfloat16,
    device_map="auto"
)
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen2.5-3B")
model = PeftModel.from_pretrained(base_model, "path/to/model/directory")

Parameters¶

class agilerl.algorithms.dpo.DPO(*args: Any, **kwargs: Any)¶

The DPO algorithm class. DPO paper: https://arxiv.org/pdf/2305.18290.

Parameters:

pad_token_id (int) – Pad token id
pad_token (str) – Pad token
model_name (str, optional) – Model name
actor_network (PreTrainedModelProtocol) – HuggingFace LLM
model_config (dict[str, Any] | None) – Model configuration, to be used when creating the model from a name or path.
hp_config (HyperparameterConfig, optional) – RL hyperparameter mutation configuration, defaults to None, whereby algorithm mutations are disabled.
index (int, optional) – Index to keep track of object instance during tournament selection and mutation, defaults to 0
batch_size (int, optional) – Batch size for training, defaults to 16
lr (float, optional) – Learning rate, defaults to 0.000005
beta (float, optional) – DPO beta parameter, defaults to 0.1
nll_alpha (float, optional) – Weight for the NLL loss on chosen responses (DPO + NLL), defaults to 1.0. Set to 0 to disable the NLL term entirely.
max_grad_norm (float, optional) – Maximum gradient norm, defaults to 0.1
update_epochs (int, optional) – Number of update epochs, defaults to 1
calc_position_embeddings (bool, optional) – Flag to indicate if position embeddings should be calculated, defaults to True
micro_batch_size_per_gpu (int, optional) – Micro batch size per GPU, defaults to None
device (str, optional) – Device for accelerated computing, ‘cpu’ or ‘cuda’, defaults to ‘cpu’
lora_config (LoraConfig, optional) – Config for LoRA, defaults to None
accelerator (accelerate.Accelerator(), optional) – Accelerator for distributed computing, defaults to None
wrap (bool, optional) – Wrap models for distributed training upon creation, defaults to True
clone (bool, optional) – Flag to indicate if the instantiation is a cloning, defaults to False
seed (int, optional) – Seed for the random number generator, defaults to 42
gradient_checkpointing (bool, optional) – Flag to indicate if gradient checkpointing should be used, defaults to True
torch_compiler (str | None, optional) – Torch compile mode (e.g. 'default'), defaults to None
use_liger_loss (bool, optional) – Use Liger kernel for memory-efficient loss computation. Defaults to False. Pass True to opt in (requires liger-kernel to be installed; warns and falls back to False otherwise). When training=False the standard path is always used regardless of this flag.
use_separate_reference_adapter (bool, optional) – Keep a dedicated reference LoRA adapter whose weights are frozen snapshots of the actor used for the DPO log-probability baseline. When False the reference log-probs are obtained by disabling the actor adapter at inference time. Defaults to True.

clean_up() → None¶: Clean up the algorithm.

clone(index: int | None = None, wrap: bool = True) → Self¶

Create a clone of the algorithm.

Parameters:

index (int | None, optional) – The index of the clone, defaults to None
wrap (bool, optional) – If True, wrap the models in the clone with the accelerator, defaults to False

Returns:

A clone of the algorithm

Return type:

EvolvableAlgorithm

static copy_attributes(agent: EvolvableAlgorithm, clone: EvolvableAlgorithm) → EvolvableAlgorithm¶

Copy the non-evolvable attributes of the algorithm to a clone.

Parameters:: clone (EvolvableAlgorithm) – The clone of the algorithm.
Returns:: The clone of the algorithm.
Return type:: EvolvableAlgorithm

evolvable_attributes(networks_only: bool = False) → dict[str, EvolvableModuleProtocol | ModuleDictProtocol | Optimizer | dict[str, Optimizer] | OptimizerWrapperProtocol]¶

Return the attributes related to the evolvable networks in the algorithm. Includes attributes that are either EvolvableModule or ModuleDict objects, as well as the optimizers associated with the networks.

Parameters:: networks_only (bool, optional) – If True, only include evolvable networks, defaults to False
Returns:: A dictionary of network attributes.
Return type:: dict[str, Any]

get_action(obs: list[ReasoningPrompts] | ReasoningPrompts, *args: Any, **kwargs: Any) → tuple[list[Tensor], list[Tensor]]¶

Return the action of the agent.

Parameters:

obs (LLMObsType) – The observation of the agent
args – Additional arguments (unused; for base contract compatibility)
kwargs – Additional keyword arguments (e.g. training; unused)

Returns:

The action of the agent

Return type:

tuple[list[torch.Tensor], list[torch.Tensor]]

Return the dimension of the action space as it pertains to the underlying networks (i.e. the output size of the networks).

Parameters:: action_space (spaces.Space or list[spaces.Space].) – The action space of the environment.
Returns:: The dimension of the action space.
Return type:: int.

get_lr_names() → list[str]¶: Return the learning rates of the algorithm.

get_policy() → EvolvableModuleProtocol¶: Return the policy network of the algorithm.

Return the dimension of the state space as it pertains to the underlying networks (i.e. the input size of the networks).

Parameters:: observation_space (spaces.Space or list[spaces.Space].) – The observation space of the environment.
Returns:: The dimension of the state space.
Return type:: tuple[int, …].

property index: int¶: Return the index of the algorithm.

static inspect_attributes(agent: EvolvableAlgorithm, input_args_only: bool = False) → dict[str, Any]¶

Inspect and retrieve the attributes of the current object, excluding attributes related to the underlying evolvable networks (i.e. EvolvableModule, torch.optim.Optimizer) and with an option to include only the attributes that are input arguments to the constructor.

Parameters:: input_args_only (bool) – If True, only include attributes that are input arguments to the constructor. Defaults to False.
Returns:: A dictionary of attribute names and their values.
Return type:: dict[str, Any]

Update agent network parameters to learn from preference data.

Parameters:

experiences (tuple[torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor, torch.Tensor]) – Batched chosen_input_ids, rejected_input_ids, chosen_attention_mask, rejected_attention_mask and rewards
training (bool) – Whether the agent is training or not

Returns:

Dict with keys mean_loss, mean_chosen_reward, mean_rejected_reward.

Return type:

dict[str, float]

classmethod load(path: str, device: str | device = 'cpu', accelerator: Accelerator | None = None) → None¶

Load an algorithm from a checkpoint.

Parameters:

path (string) – Location to load checkpoint from.
device (str, optional) – Device to load the algorithm on, defaults to ‘cpu’
accelerator (Accelerator | None, optional) – Accelerator object for distributed computing, defaults to None

Returns:

An instance of the algorithm

Return type:

RLAlgorithm

load_checkpoint(path: str, load_optimizer: bool = False, overwrite_reference_adapter: bool = False, overwrite_critic_adapter: bool = True, merge_lora_configs: bool = False) → None¶

Load adapter weights and algorithm state from a checkpoint directory.

Adapter roles restored on load:

actor — the trained policy. Always loaded.

reference — the fixed policy used for KL / comparison. The checkpoint’s actor adapter is copied onto reference so that SFT -> DPO -> GRPO chains work out of the box: the stage-N actor becomes the stage-N+1 reference.

critic — optional value head. Loaded from disk if a critic/ adapter is present, else copied from actor, else left as the live fresh LoRA init.

LoRA config reconciliation: when the checkpoint’s config and the live algorithm’s config disagree, loading fails fast by default. Pass merge_lora_configs=True to merge them for compatibility:

r (rank) -> max(current, checkpoint); the smaller side’s weights are padded into the top-left rank slice of the larger adapter (see _pad_adapter_state_to_live_shape()).

target_modules / modules_to_save -> union.

Any other mismatched field -> current value wins, with a warning.

Any adapter whose live config ends up differing from the selected target config is rebuilt via _reconfigure_adapters_to_match() before weights are loaded, so tensors always land in the correct shape.

No DeepSpeed:

lora_only=T, load_optimizer=T -> PEFT adapter load + optimizer
state from attributes.pt

lora_only=T, load_optimizer=F -> PEFT adapter load only lora_only=F, load_optimizer=T -> torch load of actor +

optimizer from attributes.pt

lora_only=F, load_optimizer=F -> torch load of actor only

DeepSpeed:

lora_only=T, load_optimizer=T -> DeepSpeed engine load from
<path>/save_checkpoint

lora_only=T, load_optimizer=F -> PEFT adapter load lora_only=F, load_optimizer=T -> DeepSpeed engine load from

<path>/save_checkpoint

lora_only=F, load_optimizer=F -> actor.load_state_dict(...)
from attributes.pt

When load_optimizer=True but the checkpoint contains no optimizer state (e.g. it was saved with save_optimizer=False), a UserWarning is emitted and a freshly-initialised optimizer is used.

Parameters:

path (str) – Directory containing a checkpoint written by save_checkpoint().
load_optimizer (bool) – If True (default) also load the optimizer and LR scheduler state so training can resume. On DeepSpeed ZeRO ≥ 2 this reads a sharded checkpoint from <path>/save_checkpoint; otherwise optimizer state is read from attributes.pt.
merge_lora_configs (bool) – If True, allow loading checkpoints whose LoRA config differs from the live agent by reconciling them. If False (default), mismatched LoRA configs raise ValueError.

property mut: Any¶: Return the mutation object of the algorithm.

mutation_hook() → None¶: Execute the hooks registered with the algorithm.

Create a population of algorithms.

Parameters:: size (int.) – The size of the population.
Returns:: A list of algorithms.
Return type:: list[EvolvableAlgorithm].

Preprocess observations (dummy) for forward pass through neural network.

Parameters:: observations (numpy.ndarray[float] or dict[str, numpy.ndarray[float]]) – Observations of environment
Returns:: Preprocessed observations
Return type:: torch.Tensor[float] or dict[str, torch.Tensor[float]] or tuple[torch.Tensor[float], …]

recompile() → None¶

Recompile evolvable modules with torch.compile.

Iterates over evolvable_attributes and compiles each one. Skipped when DeepSpeed is active because DeepSpeedEngine is not compatible with OptimizedModule wrapping.

register_mutation_hook(hook: Callable) → None¶

Parameters:: hook (Callable) – The hook to be executed after mutation.

register_network_group(group: NetworkGroup) → None¶

Set the evaluation network for the algorithm.

Parameters:: name (str) – The name of the evaluation network.

reinit_optimizers(optimizer: OptimizerConfig | None = None) → None¶

Reinitialize the optimizers of an algorithm. If no optimizer is passed, all optimizers are reinitialized.

Parameters:: optimizer (OptimizerConfig | None, optional) – The optimizer to reinitialize, defaults to None, in which case all optimizers are reinitialized.

save_checkpoint(path: str, lora_only: bool = True, save_optimizer: bool = True, **kwargs: Any) → None¶

Save adapter weights and algorithm state to a directory.

AgileRL never persists base-model weights when lora_only=True for LLM algorithms: a checkpoint is a directory containing

<adapter>/adapter_model.safetensors + adapter_config.json — one subdirectory per adapter in selected_adapters (always actor, plus reference / critic when those adapters are configured). Written only when lora_only=True.

attributes.pt — algorithm hyperparameters, plus (optionally) the actor state dict and/or optimizer state dict depending on the cell below. Always present.

save_checkpoint/ — DeepSpeed ZeRO ≥ 2 sharded-checkpoint output. Present only when an Accelerator is attached and save_optimizer=True.

Behaviour per cell of the (lora_only, save_optimizer, deepspeed) grid:

Plain (no accelerator):

lora_only=T, save_optimizer=T -> PEFT adapter dirs on disk +
optimizer state in attributes.pt

lora_only=T, save_optimizer=F -> PEFT adapter dirs only lora_only=F, save_optimizer=T -> full actor state_dict +

optimizer state in attributes.pt

lora_only=F, save_optimizer=F -> full actor state_dict in attributes.pt

DeepSpeed:

lora_only=T, save_optimizer=T -> engine tag dir (frozen params
excluded) + PEFT adapter dirs

lora_only=T, save_optimizer=F -> PEFT adapter dirs only lora_only=F, save_optimizer=T -> engine tag dir (frozen params

included)

lora_only=F, save_optimizer=F -> gathered (ZeRO-3 aware) actor
state_dict injected into attributes.pt

Parameters:

path (str) – Directory to write the checkpoint into.
lora_only (bool) – If True (default) only adapter weights are written to disk via save_pretrained; the base model is shared across checkpoints and not serialised. If False, the full actor state dict is persisted (into attributes.pt on the plain path, or into the DeepSpeed engine’s tag dir / gathered dict on the distributed path).
save_optimizer (bool) – If True (default) also persist the optimizer and LR scheduler state so training can resume. On DeepSpeed ZeRO ≥ 2 this writes a sharded checkpoint into <path>/save_checkpoint; otherwise optimizer state is included in attributes.pt.

select_adapter(adapter_name: str) → None¶

Temporarily switch adapter; restores the actor adapter on exit.

Parameters:: adapter_name (str) – Name of the adapter to activate (“actor”, “critic”, “reference”).

set_reference_policy(reference_update_tracker: int) → None¶

Update the reference policy when the reference policy update tracker is greater than the current reference policy update tracker.

Parameters:: reference_update_tracker (int) – The reference policy update tracker

set_training_mode(training: bool) → None¶

Set the training mode of the algorithm.

Parameters:: training (bool) – If True, set the algorithm to training mode.

test(env: PreferenceGym, loop: int = 1, *args: Any, **kwargs: Any) → np.ndarray¶

Return the fitness (test) score of the agent.

Parameters:

env (PreferenceGym environment) – The environment to be tested in
loop (int, optional) – Number of testing loops/episodes to complete. The returned score is the mean. Defaults to 1

Returns:

Mean test score (numpy array)

Return type:

np.ndarray

Move experiences to the device.

Parameters:: experiences (tuple[torch.Tensor[float], ...]) – Experiences to move to device
Returns:: Experiences on the device
Return type:: tuple[torch.Tensor[float], …]

unwrap_models() → None¶: Unwraps the models in the algorithm from the accelerator.

Update the learning rate of the optimizer.

Parameters:

optimizer (Optimizer) – Optimizer
lr (float | tuple[float, float]) – Learning rate value, or actor/critic pair.
accelerator (Accelerator | None) – Accelerator
scheduler_config (CosineLRScheduleConfig | None) – Scheduler configuration

Returns:

Tuple of accelerator and scheduler

Returns:

Accelerator

use_adapter(adapter_name: str) → None¶

Switch the active PEFT adapter, handling all side-effects.

For “reference”: switches adapter and freezes reference params (never trained). For all others: switches adapter and restores requires_grad=True on all training adapter LoRA params so that DeepSpeed ZeRO-2 gradient bucket hooks keep firing correctly.

Parameters:: adapter_name (str) – Name of the adapter to activate (“actor”, “critic”, “reference”).

wrap_models() → None¶: Wrap the models in the accelerator, DeepSpeed objects must be wrapped at the same time, not individually.