vendor/rllm/scripts/train/simple_math.sh

#!/bin/bash
set -x

# Warning: Export VLLM_ATTENTION_BACKEND on every machine before starting Ray cluster.
# vLLM without XFORMERS will results in CUDA errors.
export VLLM_ATTENTION_BACKEND=FLASH_ATTN
export VLLM_USE_V1=1

# Parse command line arguments
while [[ $# -gt 0 ]]; do
    case $1 in
        --model)
            MODEL_PATH="$2"
            shift 2
            ;;
        *)
            break
            ;;
    esac
done

# Set default model path if not provided
if [ -z "$MODEL_PATH" ]; then
    MODEL_PATH="Qwen/Qwen3-4B"
fi

# Train over a single node, 8 A100-80GB GPUs.
python3 -m verl.trainer.main_ppo \
    algorithm.adv_estimator=grpo \
    data.train_files=$HOME/rllm-internal/data/math_train.parquet \
    data.val_files=$HOME/rllm-internal/data/math.parquet \
    data.train_batch_size=64 \
    data.val_batch_size=512 \
    data.max_prompt_length=2048 \
    data.max_response_length=2048 \
    data.return_raw_chat=True \
    data.truncation=error \
    actor_rollout_ref.model.path=$MODEL_PATH  \
    actor_rollout_ref.actor.optim.lr=1e-6 \
    actor_rollout_ref.model.use_remove_padding=True \
    actor_rollout_ref.actor.ppo_mini_batch_size=16 \
    actor_rollout_ref.actor.use_dynamic_bsz=True \
    actor_rollout_ref.actor.ppo_max_token_len_per_gpu=24000 \
    actor_rollout_ref.actor.use_kl_loss=True \
    actor_rollout_ref.actor.kl_loss_coef=0.001 \
    actor_rollout_ref.actor.kl_loss_type=low_var_kl \
    actor_rollout_ref.actor.ulysses_sequence_parallel_size=1 \
    actor_rollout_ref.model.enable_gradient_checkpointing=True \
    actor_rollout_ref.actor.fsdp_config.param_offload=False \
    actor_rollout_ref.actor.fsdp_config.optimizer_offload=False \
    actor_rollout_ref.rollout.tensor_model_parallel_size=1 \
    actor_rollout_ref.rollout.name=vllm \
    actor_rollout_ref.rollout.temperature=0.6 \
    actor_rollout_ref.rollout.mode="async" \
    actor_rollout_ref.rollout.chat_scheduler=verl.schedulers.naive_chat_scheduler.NaiveChatCompletionScheduler \
    actor_rollout_ref.rollout.val_kwargs.temperature=0.6 \
    actor_rollout_ref.rollout.gpu_memory_utilization=0.85 \
    actor_rollout_ref.rollout.n=4 \
    actor_rollout_ref.rollout.val_kwargs.n=1 \
    actor_rollout_ref.rollout.val_kwargs.top_p=0.95 \
    actor_rollout_ref.rollout.enforce_eager=False \
    actor_rollout_ref.ref.fsdp_config.param_offload=True \
    algorithm.kl_ctrl.kl_coef=0.001 \
    trainer.critic_warmup=0 \
    trainer.logger=['console','wandb'] \
    trainer.project_name='deepscaler' \
    trainer.experiment_name='deepscaler-math-debug' \
    trainer.val_before_train=True \
    trainer.n_gpus_per_node=8 \
    trainer.nnodes=1 \
    trainer.save_freq=10 \
    trainer.test_freq=10 \
    trainer.default_hdfs_dir=null \
    trainer.total_epochs=30 "${@:1}"