open-thought/reasoning-gym
1
0
Fork 0
mirror of https://github.com/open-thought/reasoning-gym.git synced 2026-04-28 17:29:39 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

add grpo launch script

Browse source
This commit is contained in:
Andreas Koepf 2025-02-17 20:52:03 +00:00
parent b1e07b7782
commit c40069e616
4 changed files with 212 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

30
examples/veRL/train_ppo.sh Executable file
View file

@ -0,0 +1,30 @@
#!/bin/bash
python3 -u main_ppo_custom_reward.py \
data.train_files=$DATA_DIR/train.parquet \
data.val_files=$DATA_DIR/test.parquet \
data.train_batch_size=256 \
data.val_batch_size=1312 \
data.max_prompt_length=256 \
data.max_response_length=1024 \
actor_rollout_ref.model.path=$BASE_MODEL \
actor_rollout_ref.actor.optim.lr=1e-6 \
actor_rollout_ref.actor.ppo_mini_batch_size=128 \
actor_rollout_ref.actor.ppo_micro_batch_size=8 \
actor_rollout_ref.rollout.log_prob_micro_batch_size=8 \
actor_rollout_ref.rollout.tensor_model_parallel_size=$ROLLOUT_TP_SIZE \
actor_rollout_ref.rollout.gpu_memory_utilization=0.4 \
actor_rollout_ref.ref.log_prob_micro_batch_size=4 \
critic.optim.lr=1e-5 \
critic.model.path=$BASE_MODEL \
critic.ppo_micro_batch_size=8 \
algorithm.kl_ctrl.kl_coef=0.001 \
trainer.logger=['wandb'] \
+trainer.val_before_train=False \
trainer.default_hdfs_dir=null \
trainer.n_gpus_per_node=$N_GPUS \
trainer.nnodes=1 \
trainer.save_freq=100 \
trainer.test_freq=100 \
trainer.project_name='verl_chain_sum_ppo' \
trainer.experiment_name=$EXPERIMENT_NAME \
trainer.total_epochs=15 2>&1 | tee verl_output.log