Agent-R1/run_rpp.sh at main · MultiturnRL/Agent-R1 · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
export VLLM_ATTENTION_BACKEND=XFORMERS
export BASE_MODEL='Qwen/Qwen2.5-1.5B-Instruct'
export PROJECT_NAME='hotpotqa_qwen2.5-1.5b-instruct'
export EXPERIMENT_NAME=reinforce_plus_plus
export HYDRA_FULL_ERROR=1
export CUDA_LAUNCH_BLOCKING=1

python3 -m agent_r1.src.main_agent \
    data.train_files=./data/hotpotqa/train.parquet \
    data.val_files=./data/hotpotqa/validation.parquet \
    data.train_batch_size=128 \
    data.max_prompt_length=4096 \
    data.max_response_length=4096 \
    data.max_start_length=4096 \
    data.max_tool_response_length=4096 \
    actor_rollout_ref.model.path=$BASE_MODEL \
    actor_rollout_ref.actor.optim.lr=1e-6 \
    actor_rollout_ref.model.use_remove_padding=True \
    actor_rollout_ref.actor.ppo_mini_batch_size=64 \
    actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=2 \
    actor_rollout_ref.model.enable_gradient_checkpointing=True \
    actor_rollout_ref.actor.fsdp_config.param_offload=False \
    actor_rollout_ref.actor.fsdp_config.optimizer_offload=False \
    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=2 \
    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \
    actor_rollout_ref.rollout.name=vllm \
    actor_rollout_ref.rollout.gpu_memory_utilization=0.5 \
    actor_rollout_ref.ref.log_prob_micro_batch_size_per_gpu=2 \
    actor_rollout_ref.ref.fsdp_config.param_offload=True \
    critic.optim.lr=1e-5 \
    critic.model.use_remove_padding=True \
    critic.model.path=$BASE_MODEL \
    critic.model.enable_gradient_checkpointing=True \
    critic.ppo_micro_batch_size_per_gpu=2 \
    critic.model.fsdp_config.param_offload=False \
    critic.model.fsdp_config.optimizer_offload=False \
    algorithm.adv_estimator=reinforce_plus_plus \
    algorithm.kl_ctrl.kl_coef=0.001 \
    trainer.critic_warmup=3 \
    trainer.logger=['console','wandb'] \
    trainer.project_name=$PROJECT_NAME \
    trainer.experiment_name=$EXPERIMENT_NAME \
    trainer.n_gpus_per_node=8 \
    trainer.nnodes=1 \
    trainer.save_freq=-1 \
    trainer.test_freq=10 \
    trainer.total_epochs=1 \
    trainer.val_before_train=True \
    tool.env='search' $@