minor (f52159df) · Commits · 张泽凯 / VAGEN

vagen/examples/debug_sokoban_vision/run.sh

+3 −3

Original line number	Diff line number	Diff line
		@@ -24,7 +24,7 @@ python3 -m vagen.trainer.main_ppo \
		data.max_trajectory_length=1800 \
		data.image_key=images \
		data.truncation=left \
		actor_rollout_ref.model.path=Qwen/Qwen2.5-0.5B-Instruct \
		actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
		actor_rollout_ref.actor.optim.lr=1e-6 \
		actor_rollout_ref.model.use_remove_padding=True \
		actor_rollout_ref.actor.ppo_mini_batch_size=32 \
		@@ -49,7 +49,7 @@ python3 -m vagen.trainer.main_ppo \
		actor_rollout_ref.rollout.temperature=0.7 \
		critic.optim.lr=1e-5 \
		critic.model.use_remove_padding=True \
		critic.model.path=Qwen/Qwen2.5-0.5B-Instruct \
		critic.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
		critic.model.enable_gradient_checkpointing=True \
		critic.ppo_micro_batch_size_per_gpu=1 \
		critic.model.fsdp_config.param_offload=False \
		@@ -58,7 +58,7 @@ python3 -m vagen.trainer.main_ppo \
		trainer.critic_warmup=0 \
		trainer.logger=['console','wandb'] \
		trainer.project_name='vagen_debug' \
		trainer.experiment_name='grpo_mask_loss_sokoban_text_debug' \
		trainer.experiment_name='grpo_mask_loss_sokoban_vision_debug' \
		trainer.n_gpus_per_node=2 \
		trainer.nnodes=1 \
		trainer.save_freq=100 \