Commit 5faaec07 authored by jameskrw's avatar jameskrw
Browse files

minor

parent fe3c7ce2

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

+4 −3
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=648 \

    data.max_trajectory_length=2600 \

    data.max_trajectory_length=3000 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -42,9 +42,9 @@ python3 -m vagen.trainer.main_ppo \ 
    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=2 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \

    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 \
@@ -79,4 +79,5 @@ python3 -m vagen.trainer.main_ppo \ 
    trainer.val_generations_to_log_to_wandb=8 \

    rollout_manager.n_trajectory=1 \

    rollout_manager.use_service=True \

    rollout_manager.base_url="http://localhost:5000" \

    2>&1 | tee $EXPERIMENT_NAME.log

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

+4 −3
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=648 \

    data.max_trajectory_length=2600 \

    data.max_trajectory_length=3000 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -42,9 +42,9 @@ python3 -m vagen.trainer.main_ppo \ 
    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=2 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \

    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 \
@@ -79,4 +79,5 @@ python3 -m vagen.trainer.main_ppo \ 
    trainer.val_generations_to_log_to_wandb=8 \

    rollout_manager.n_trajectory=1 \

    rollout_manager.use_service=True \

    rollout_manager.base_url="http://localhost:5000" \

    2>&1 | tee $EXPERIMENT_NAME.log

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

+4 −3
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=648 \

    data.max_trajectory_length=2600 \

    data.max_trajectory_length=3000 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -42,9 +42,9 @@ python3 -m vagen.trainer.main_ppo \ 
    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=2 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \

    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 \
@@ -79,4 +79,5 @@ python3 -m vagen.trainer.main_ppo \ 
    trainer.val_generations_to_log_to_wandb=8 \

    rollout_manager.n_trajectory=1 \

    rollout_manager.use_service=True \

    rollout_manager.base_url="http://localhost:5000" \

    2>&1 | tee $EXPERIMENT_NAME.log

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

+4 −3
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=648 \

    data.max_trajectory_length=2600 \

    data.max_trajectory_length=3000 \

    data.image_key=images \

    data.truncation=error \

    actor_rollout_ref.model.path=Qwen/Qwen2.5-VL-3B-Instruct \
@@ -42,9 +42,9 @@ python3 -m vagen.trainer.main_ppo \ 
    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=2 \

    actor_rollout_ref.rollout.tensor_model_parallel_size=4 \

    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 \
@@ -79,4 +79,5 @@ python3 -m vagen.trainer.main_ppo \ 
    trainer.val_generations_to_log_to_wandb=8 \

    rollout_manager.n_trajectory=1 \

    rollout_manager.use_service=True \

    rollout_manager.base_url="http://localhost:5000" \

    2>&1 | tee $EXPERIMENT_NAME.log

CRA Git | Maintained and supported by SUSTech CRA and CCSE