mirror of
https://github.com/volcengine/verl.git
synced 2025-10-20 13:43:50 +08:00
31 lines
1.3 KiB
Bash
31 lines
1.3 KiB
Bash
set -x
|
|
|
|
python3 -m verl.trainer.main_ppo --config-path=./config --config-name='ppo_megatron_trainer'\
|
|
data.train_files=$HOME/data/gsm8k/train.parquet \
|
|
data.val_files=$HOME/data/gsm8k/test.parquet \
|
|
data.train_batch_size=1024 \
|
|
data.val_batch_size=1312 \
|
|
data.max_prompt_length=512 \
|
|
data.max_response_length=512 \
|
|
actor_rollout_ref.model.path=deepseek-ai/deepseek-coder-6.7b-instruct \
|
|
actor_rollout_ref.actor.optim.lr=2e-6 \
|
|
actor_rollout_ref.actor.ppo_mini_batch_size=256 \
|
|
actor_rollout_ref.actor.ppo_micro_batch_size=64 \
|
|
actor_rollout_ref.rollout.log_prob_micro_batch_size=64 \
|
|
actor_rollout_ref.rollout.tensor_model_parallel_size=2 \
|
|
actor_rollout_ref.rollout.name=vllm \
|
|
actor_rollout_ref.rollout.gpu_memory_utilization=0.5 \
|
|
actor_rollout_ref.ref.log_prob_micro_batch_size=128 \
|
|
critic.optim.lr=2e-5 \
|
|
critic.model.path=deepseek-ai/deepseek-coder-6.7b-instruct \
|
|
critic.model.enable_gradient_checkpointing=False \
|
|
critic.ppo_micro_batch_size=64 \
|
|
algorithm.kl_ctrl.kl_coef=0.001 \
|
|
trainer.critic_warmup=0 \
|
|
trainer.logger=['console','tracking'] \
|
|
trainer.project_name='verl_megatron_gsm8k_examples' \
|
|
trainer.experiment_name='deepseek_llm_7b_function_rm' \
|
|
trainer.n_gpus_per_node=8 \
|
|
trainer.nnodes=1 \
|
|
trainer.save_freq=-1 \
|
|
trainer.total_epochs=15 |