wehub-resource-sync
44 lines
1.5 KiB
Bash
44 lines
1.5 KiB
Bash
# 8*80G GPU
|
|
# GSPO https://arxiv.org/pdf/2507.18071
|
|
# hyperparameter
|
|
# - epsilon = 3e-4 from paper serction 5.1
|
|
# - epsilon_high = 4e-4 from paper serction 5.1
|
|
# - steps_per_generation = 32 (= 4 * gradient_accumulation_steps): paper section 5.1 partitions each batch of rollout data into four minibatches for gradient updates; in swift HF GRPO, steps_per_generation counts micro-batches, so it must be multiplied by gradient_accumulation_steps
|
|
# - beta = 0: zero kl regularization https://github.com/volcengine/verl/pull/2775#issuecomment-3131807306
|
|
|
|
CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 \
|
|
NPROC_PER_NODE=8 \
|
|
swift rlhf \
|
|
--rlhf_type grpo \
|
|
--model Qwen/Qwen2.5-7B-Instruct \
|
|
--dataset AI-MO/NuminaMath-TIR#10000 \
|
|
--load_from_cache_file true \
|
|
--torch_dtype bfloat16 \
|
|
--beta 0.0 \
|
|
--epsilon 3e-4 \
|
|
--epsilon_high 4e-4 \
|
|
--steps_per_generation 32 \
|
|
--importance_sampling_level sequence \
|
|
--num_train_epochs 1 \
|
|
--per_device_train_batch_size 2 \
|
|
--gradient_accumulation_steps 8 \
|
|
--num_generations 16 \
|
|
--tuner_type full \
|
|
--reward_funcs accuracy \
|
|
--use_vllm true \
|
|
--vllm_mode colocate \
|
|
--vllm_gpu_memory_utilization 0.6 \
|
|
--vllm_max_model_len 16384 \
|
|
--max_completion_length 8192 \
|
|
--offload_optimizer true \
|
|
--offload_model true \
|
|
--sleep_level 1 \
|
|
--save_steps 1000 \
|
|
--learning_rate 1e-6 \
|
|
--save_total_limit 2 \
|
|
--logging_steps 5 \
|
|
--warmup_ratio 0.05 \
|
|
--dataloader_num_workers 4 \
|
|
--deepspeed zero3 \
|
|
--log_completions true
|