legacy_train.sh

#!/bin/bash

# This script is only maintained on the CI for backward compatibility testing.
# You will need to run the following Python script as a companion:
#   python legacy_calc_agent.py

set -ex

export N_GPUS=1
export BASE_MODEL=Qwen/Qwen2.5-1.5B-Instruct
export DATA_DIR=data
export ROLLOUT_TP_SIZE=1
export EXPERIMENT_NAME="calc_x_$(date +%Y%m%d%H%M%S)"
export PROJECT_NAME=AgentLightningCI
echo "project_name=${PROJECT_NAME}" >> $GITHUB_OUTPUT
echo "run_name=${EXPERIMENT_NAME}" >> $GITHUB_OUTPUT

PYTHONUNBUFFERED=1 python -m agentlightning.verl \
    algorithm.adv_estimator=grpo \
    data.train_files=${DATA_DIR}/train.parquet \
    data.val_files=${DATA_DIR}/test_mini.parquet \
    actor_rollout_ref.rollout.tensor_model_parallel_size=$ROLLOUT_TP_SIZE \
    trainer.n_gpus_per_node=${N_GPUS} \
    data.train_batch_size=32 \
    actor_rollout_ref.rollout.n=4 \
    actor_rollout_ref.actor.ppo_mini_batch_size=32 \
    actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=4 \
    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=4 \
    actor_rollout_ref.rollout.multi_turn.format=hermes \
    actor_rollout_ref.model.path=${BASE_MODEL} \
    data.max_prompt_length=4096 \
    data.max_response_length=2048 \
    data.truncation='error' \
    trainer.val_before_train=True \
    actor_rollout_ref.actor.optim.lr=1e-6 \
    actor_rollout_ref.model.use_remove_padding=True \
    actor_rollout_ref.actor.use_kl_loss=False \
    actor_rollout_ref.actor.kl_loss_coef=0.000 \
    actor_rollout_ref.actor.entropy_coeff=0 \
    actor_rollout_ref.actor.clip_ratio_low=0.2 \
    actor_rollout_ref.actor.clip_ratio_high=0.3 \
    actor_rollout_ref.model.enable_gradient_checkpointing=True \
    actor_rollout_ref.actor.fsdp_config.param_offload=True \
    actor_rollout_ref.actor.fsdp_config.optimizer_offload=True \
    actor_rollout_ref.rollout.name=vllm \
    actor_rollout_ref.rollout.gpu_memory_utilization=0.8 \
    actor_rollout_ref.ref.log_prob_micro_batch_size_per_gpu=8 \
    actor_rollout_ref.ref.fsdp_config.param_offload=True \
    algorithm.use_kl_in_reward=False \
    trainer.critic_warmup=0 \
    trainer.logger=['console','wandb'] \
    trainer.project_name=${PROJECT_NAME} \
    trainer.experiment_name=${EXPERIMENT_NAME} \
    trainer.nnodes=1 \
    trainer.test_freq=6 \
    trainer.total_epochs=1 \
    trainer.total_training_steps=6 $@