Commit e7e7077d authored by YaningGao's avatar YaningGao
Browse files

Merge branch 'main' of github.com:RAGEN-AI/vagen into prompt_dev

parents a593ebc1 ba3b8410

scripts/exps/format/frozenlake/free_think/run.sh

+6 −6
Original line number Diff line number Diff line
@@ -26,8 +26,8 @@ python3 -m vagen.trainer.main_ppo \ 
    data.val_files=data/$EXPERIMENT_NAME/test.parquet \

    data.train_batch_size=64 \

    data.max_prompt_length=1024 \

    data.max_response_length=256 \

    data.max_trajectory_length=3072 \

    data.max_response_length=150 \

    data.max_trajectory_length=2200 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -39,10 +39,10 @@ python3 -m vagen.trainer.main_ppo \ 
    actor_rollout_ref.actor.kl_loss_coef=0.001 \

    actor_rollout_ref.actor.kl_loss_type=mse \

    actor_rollout_ref.model.enable_gradient_checkpointing=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=True \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=False \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=False \

    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=1 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=2 \

    actor_rollout_ref.rollout.name=vllm \

    actor_rollout_ref.rollout.gpu_memory_utilization=0.1 \

    actor_rollout_ref.rollout.enable_chunked_prefill=False \
@@ -67,7 +67,7 @@ python3 -m vagen.trainer.main_ppo \ 
    trainer.experiment_name=$EXPERIMENT_NAME \

    trainer.n_gpus_per_node=4 \

    trainer.nnodes=1 \

    trainer.save_freq=70 \

    trainer.save_freq=90 \

    trainer.test_freq=20 \

    trainer.total_training_steps=200 \

    rollout_manager.max_turns=3 \

scripts/exps/format/frozenlake/grounding/run.sh

+6 −6
Original line number Diff line number Diff line
@@ -26,8 +26,8 @@ python3 -m vagen.trainer.main_ppo \ 
    data.val_files=data/$EXPERIMENT_NAME/test.parquet \

    data.train_batch_size=64 \

    data.max_prompt_length=1024 \

    data.max_response_length=256 \

    data.max_trajectory_length=3072 \

    data.max_response_length=150 \

    data.max_trajectory_length=2200 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -39,10 +39,10 @@ python3 -m vagen.trainer.main_ppo \ 
    actor_rollout_ref.actor.kl_loss_coef=0.001 \

    actor_rollout_ref.actor.kl_loss_type=mse \

    actor_rollout_ref.model.enable_gradient_checkpointing=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=True \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=False \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=False \

    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=1 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=2 \

    actor_rollout_ref.rollout.name=vllm \

    actor_rollout_ref.rollout.gpu_memory_utilization=0.1 \

    actor_rollout_ref.rollout.enable_chunked_prefill=False \
@@ -67,7 +67,7 @@ python3 -m vagen.trainer.main_ppo \ 
    trainer.experiment_name=$EXPERIMENT_NAME \

    trainer.n_gpus_per_node=4 \

    trainer.nnodes=1 \

    trainer.save_freq=70 \

    trainer.save_freq=90 \

    trainer.test_freq=20 \

    trainer.total_training_steps=200 \

    rollout_manager.max_turns=3 \

scripts/exps/format/frozenlake/grounding_worldmodeling/run.sh

+6 −6
Original line number Diff line number Diff line
@@ -26,8 +26,8 @@ python3 -m vagen.trainer.main_ppo \ 
    data.val_files=data/$EXPERIMENT_NAME/test.parquet \

    data.train_batch_size=64 \

    data.max_prompt_length=1024 \

    data.max_response_length=256 \

    data.max_trajectory_length=3072 \

    data.max_response_length=150 \

    data.max_trajectory_length=2200 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -39,10 +39,10 @@ python3 -m vagen.trainer.main_ppo \ 
    actor_rollout_ref.actor.kl_loss_coef=0.001 \

    actor_rollout_ref.actor.kl_loss_type=mse \

    actor_rollout_ref.model.enable_gradient_checkpointing=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=True \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=False \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=False \

    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=1 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=2 \

    actor_rollout_ref.rollout.name=vllm \

    actor_rollout_ref.rollout.gpu_memory_utilization=0.1 \

    actor_rollout_ref.rollout.enable_chunked_prefill=False \
@@ -67,7 +67,7 @@ python3 -m vagen.trainer.main_ppo \ 
    trainer.experiment_name=$EXPERIMENT_NAME \

    trainer.n_gpus_per_node=4 \

    trainer.nnodes=1 \

    trainer.save_freq=70 \

    trainer.save_freq=90 \

    trainer.test_freq=20 \

    trainer.total_training_steps=200 \

    rollout_manager.max_turns=3 \

scripts/exps/format/frozenlake/no_think/run.sh

+6 −6
Original line number Diff line number Diff line
@@ -26,8 +26,8 @@ python3 -m vagen.trainer.main_ppo \ 
    data.val_files=data/$EXPERIMENT_NAME/test.parquet \

    data.train_batch_size=64 \

    data.max_prompt_length=1024 \

    data.max_response_length=256 \

    data.max_trajectory_length=3072 \

    data.max_response_length=150 \

    data.max_trajectory_length=2200 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -39,10 +39,10 @@ python3 -m vagen.trainer.main_ppo \ 
    actor_rollout_ref.actor.kl_loss_coef=0.001 \

    actor_rollout_ref.actor.kl_loss_type=mse \

    actor_rollout_ref.model.enable_gradient_checkpointing=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=True \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=False \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=False \

    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=1 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=2 \

    actor_rollout_ref.rollout.name=vllm \

    actor_rollout_ref.rollout.gpu_memory_utilization=0.1 \

    actor_rollout_ref.rollout.enable_chunked_prefill=False \
@@ -67,7 +67,7 @@ python3 -m vagen.trainer.main_ppo \ 
    trainer.experiment_name=$EXPERIMENT_NAME \

    trainer.n_gpus_per_node=4 \

    trainer.nnodes=1 \

    trainer.save_freq=70 \

    trainer.save_freq=90 \

    trainer.test_freq=20 \

    trainer.total_training_steps=200 \

    rollout_manager.max_turns=3 \

scripts/exps/format/frozenlake/worldmodeling/run.sh

+6 −6
Original line number Diff line number Diff line
@@ -26,8 +26,8 @@ python3 -m vagen.trainer.main_ppo \ 
    data.val_files=data/$EXPERIMENT_NAME/test.parquet \

    data.train_batch_size=64 \

    data.max_prompt_length=1024 \

    data.max_response_length=256 \

    data.max_trajectory_length=3072 \

    data.max_response_length=150 \

    data.max_trajectory_length=2200 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -39,10 +39,10 @@ python3 -m vagen.trainer.main_ppo \ 
    actor_rollout_ref.actor.kl_loss_coef=0.001 \

    actor_rollout_ref.actor.kl_loss_type=mse \

    actor_rollout_ref.model.enable_gradient_checkpointing=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=True \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=True \

    actor_rollout_ref.actor.fsdp_config.param_offload=False \

    actor_rollout_ref.actor.fsdp_config.optimizer_offload=False \

    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=1 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=2 \

    actor_rollout_ref.rollout.name=vllm \

    actor_rollout_ref.rollout.gpu_memory_utilization=0.1 \

    actor_rollout_ref.rollout.enable_chunked_prefill=False \
@@ -67,7 +67,7 @@ python3 -m vagen.trainer.main_ppo \ 
    trainer.experiment_name=$EXPERIMENT_NAME \

    trainer.n_gpus_per_node=4 \

    trainer.nnodes=1 \

    trainer.save_freq=70 \

    trainer.save_freq=90 \

    trainer.test_freq=20 \

    trainer.total_training_steps=200 \

    rollout_manager.max_turns=3 \

CRA Git | Maintained and supported by SUSTech CRA and CCSE