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

Basic curriculum (#198)

Browse source 
* feat: Add optional curriculum support to dataset registration and creation
* docs: Add docstrings to create_curriculum() and register_dataset()
* feat: Add curriculum configuration classes for CurriculumExperiment
* feat: Add weight parameter to CurriculumAttributeConfig and use in DatasetSpec
* refactor: Simplify CurriculumAttributeConfig with "*" attribute level support
* test: Add unit tests for CurriculumExperiment class
* feat: Add from_yaml() method to CurriculumExperimentConfig with unit test
This commit is contained in:
Andreas Köpf 2025-03-07 11:22:12 +01:00 • committed by GitHub
parent cbfdf097a0
commit c69bc5d4e6
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
29 changed files with 943 additions and 63 deletions
Download patch file Download diff file Expand all files Collapse all files

39
examples/veRL/basic_curriculum/train_grpo.sh Normal file
View file

@ -0,0 +1,39 @@
#!/bin/bash
set -x
python3 -u ppo_curriculum.py \
algorithm.adv_estimator=grpo \
data.train_files=$DATA_DIR/train.parquet \
data.val_files=$DATA_DIR/test.parquet \
data.train_batch_size=512 \
data.val_batch_size=512 \
data.max_prompt_length=512 \
data.max_response_length=1024 \
actor_rollout_ref.model.path=$BASE_MODEL \
actor_rollout_ref.actor.optim.lr=1e-6 \
actor_rollout_ref.model.use_remove_padding=True \
actor_rollout_ref.actor.ppo_mini_batch_size=256 \
actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=80 \
actor_rollout_ref.actor.use_kl_loss=True \
actor_rollout_ref.actor.kl_loss_coef=0.001 \
actor_rollout_ref.actor.kl_loss_type=low_var_kl \
actor_rollout_ref.model.enable_gradient_checkpointing=True \
actor_rollout_ref.actor.fsdp_config.param_offload=False \
actor_rollout_ref.actor.fsdp_config.optimizer_offload=False \
actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=160 \
actor_rollout_ref.rollout.tensor_model_parallel_size=$ROLLOUT_TP_SIZE \
actor_rollout_ref.rollout.name=vllm \
actor_rollout_ref.rollout.gpu_memory_utilization=0.6 \
actor_rollout_ref.rollout.n=8 \
actor_rollout_ref.ref.log_prob_micro_batch_size_per_gpu=160 \
actor_rollout_ref.ref.fsdp_config.param_offload=True \
algorithm.kl_ctrl.kl_coef=0.001 \
trainer.critic_warmup=0 \
trainer.logger=['wandb'] \
trainer.project_name='verl_chain_sum_grpo' \
trainer.experiment_name=$EXPERIMENT_NAME \
trainer.n_gpus_per_node=$N_GPUS \
trainer.nnodes=1 \
trainer.save_freq=100 \
trainer.test_freq=100 \
trainer.total_epochs=15 $@ 2>&1 | tee verl_output.log