recipes/Qwen2.5-1.5B-Instruct/grpo/config_demo_code.yaml

# Model arguments
model_name_or_path: Qwen/Qwen2.5-1.5B-Instruct
model_revision: main
torch_dtype: bfloat16
attn_implementation: flash_attention_2

# Data training arguments
dataset_name: open-r1/verifiable-coding-problems-python
dataset_prompt_column: problem_statement
system_prompt: "You are a helpful AI Assistant that provides well-reasoned and detailed responses. You first think about the reasoning process as an internal monologue and then provide the user with the answer. Respond in the following format: <think>\n...\n</think>\n<answer>\n...\n</answer>"

# GRPO trainer config
beta: 0.01
bf16: true
use_vllm: true
do_eval: false
gradient_accumulation_steps: 4
gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
hub_model_id: Qwen2.5-1.5B-Open-R1-Code-GRPO
hub_strategy: every_save
learning_rate: 5.0e-06
log_completions: true
log_level: info
logging_first_step: true
logging_steps: 1
logging_strategy: steps
lr_scheduler_type: cosine_with_min_lr
lr_scheduler_kwargs:
  min_lr_rate: 0.1
max_prompt_length: 1024
max_completion_length: 2048
max_steps: 500
num_generations: 14
num_train_epochs: 1
output_dir: data/Qwen2.5-1.5B-Open-R1-Code-GRPO
overwrite_output_dir: true
per_device_train_batch_size: 16
push_to_hub: true
report_to:
- wandb
reward_funcs:
- code
- format
reward_weights:
- 1.0
- 0.1
save_strategy: "steps"
save_steps: 50
save_total_limit: 1
seed: 42
temperature: 1.0
warmup_ratio: 0.03