update

examples/tutorials/grpo/step_2_train_grpo_model_unsloth_ddp.py

@@ -20,11 +20,11 @@ nohup torchrun --nproc_per_node=4 step_2_train_grpo_model_unsloth_ddp.py \
 --valid_dataset_size 320 \
 --lora_rank 16 \
 --num_train_epochs 1 \
---per_device_batch_size 32 \
---gradient_accumulation_steps 1 \
+--per_device_batch_size 16 \
+--gradient_accumulation_steps 2 \
 --learning_rate 5e-6 \
 --num_generations 8 \
---steps_per_generation 4 \
+--steps_per_generation 16 \
 &
 
 kill -9 `ps -aef | grep 'step_2_train_grpo_model_unsloth_ddp.py' | grep -v grep | awk '{print $2}'`