Commit 6dcabad9 authored by jameskrw's avatar jameskrw
Browse files

prompt format and experiments change

parent fcdc30b3

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

+5 −5
Original line number Diff line number Diff line
@@ -27,7 +27,7 @@ python3 -m vagen.trainer.main_ppo \ 
    data.train_batch_size=128 \

    data.max_prompt_length=1024 \

    data.max_response_length=256 \

    data.max_trajectory_length=3072 \

    data.max_trajectory_length=2400 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -39,12 +39,12 @@ 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.gpu_memory_utilization=0.2 \

    actor_rollout_ref.rollout.enable_chunked_prefill=False \

    actor_rollout_ref.rollout.enforce_eager=False \

    actor_rollout_ref.rollout.free_cache_engine=False \

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

+5 −5
Original line number Diff line number Diff line
@@ -27,7 +27,7 @@ python3 -m vagen.trainer.main_ppo \ 
    data.train_batch_size=128 \

    data.max_prompt_length=1024 \

    data.max_response_length=256 \

    data.max_trajectory_length=3072 \

    data.max_trajectory_length=2400 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -39,12 +39,12 @@ 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.gpu_memory_utilization=0.2 \

    actor_rollout_ref.rollout.enable_chunked_prefill=False \

    actor_rollout_ref.rollout.enforce_eager=False \

    actor_rollout_ref.rollout.free_cache_engine=False \

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

+2 −2
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=128 \

    data.max_prompt_length=1024 \

    data.max_response_length=128 \

    data.max_trajectory_length=2048 \

    data.max_response_length=256 \

    data.max_trajectory_length=2400 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \

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

+5 −5
Original line number Diff line number Diff line
@@ -27,7 +27,7 @@ python3 -m vagen.trainer.main_ppo \ 
    data.train_batch_size=128 \

    data.max_prompt_length=1024 \

    data.max_response_length=256 \

    data.max_trajectory_length=3072 \

    data.max_trajectory_length=2400 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -39,12 +39,12 @@ 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.gpu_memory_utilization=0.2 \

    actor_rollout_ref.rollout.enable_chunked_prefill=False \

    actor_rollout_ref.rollout.enforce_eager=False \

    actor_rollout_ref.rollout.free_cache_engine=False \

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

+5 −5
Original line number Diff line number Diff line
@@ -27,7 +27,7 @@ python3 -m vagen.trainer.main_ppo \ 
    data.train_batch_size=128 \

    data.max_prompt_length=1024 \

    data.max_response_length=256 \

    data.max_trajectory_length=3072 \

    data.max_trajectory_length=2400 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -39,12 +39,12 @@ 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.gpu_memory_utilization=0.2 \

    actor_rollout_ref.rollout.enable_chunked_prefill=False \

    actor_rollout_ref.rollout.enforce_eager=False \

    actor_rollout_ref.rollout.free_cache_engine=False \

CRA Git | Maintained and supported by SUSTech CRA and CCSE