Kashif Rasul 90a6de94c7
Revert "Weighted reward functions (#213)" (#317)
This reverts commit fbea53267b9676fc89e92c9a24c83cb23e0884d0.
2025-02-13 15:00:05 +01:00

48 lines
1.0 KiB
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: AI-MO/NuminaMath-TIR
dataset_configs:
- all
# Num processes is less by 1 as vLLM is using 1 GPU
num_processes: 7
# GRPO trainer config
bf16: true
use_vllm: true
vllm_device: auto
vllm_gpu_memory_utilization: 0.7
do_eval: true
eval_strategy: steps
eval_steps: 100
gradient_accumulation_steps: 16
gradient_checkpointing: true
gradient_checkpointing_kwargs:
use_reentrant: false
hub_model_id: Qwen2.5-1.5B-Open-R1-GRPO
hub_strategy: every_save
learning_rate: 2.0e-05
log_level: info
logging_steps: 5
logging_strategy: steps
lr_scheduler_type: cosine
max_prompt_length: 512
max_completion_length: 1024
max_steps: -1
num_generations: 7
num_train_epochs: 1
output_dir: data/Qwen2.5-1.5B-Open-R1-GRPO
overwrite_output_dir: true
per_device_eval_batch_size: 32
per_device_train_batch_size: 16
push_to_hub: true
report_to:
- wandb
save_strategy: "no"
seed: 42
warmup_ratio: 0.1