Retro-R1/test_chembl_script.sh at main · GenSI-THUAIR/Retro-R1 · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
#!/bin/bash


export_pth=./export_model/retro_qwen2.5-7b-instruct-1M/ppo_retro/global_step_190/actor
python3 -m agent_r1.src.main_agent_retro_noback     \
    data.train_files="./data/reaction_pathway_search/train_h4_10.parquet"     \
    data.val_files="./data/reaction_pathway_search/validation_chembl_1000.parquet"     \
    data.train_batch_size=128     \
    data.max_prompt_length=32768     \
    data.max_response_length=32768     \
    data.max_start_length=1024     \
    data.max_tool_response_length=4096     \
    actor_rollout_ref.model.path=$export_pth     \
    actor_rollout_ref.actor.optim.lr=1e-6     \
    actor_rollout_ref.model.use_remove_padding=True     \
    actor_rollout_ref.actor.ppo_mini_batch_size=64     \
    actor_rollout_ref.actor.ppo_micro_batch_size_per_gpu=2     \
    actor_rollout_ref.actor.use_dynamic_bsz=True     \
    actor_rollout_ref.actor.ppo_max_token_len_per_gpu=32768     \
    actor_rollout_ref.model.enable_gradient_checkpointing=True     \
    actor_rollout_ref.actor.fsdp_config.param_offload=True     \
    actor_rollout_ref.actor.fsdp_config.optimizer_offload=True     \
    actor_rollout_ref.actor.fsdp_config.fsdp_size=8     \
    actor_rollout_ref.rollout.log_prob_micro_batch_size_per_gpu=2     \
    actor_rollout_ref.rollout.tensor_model_parallel_size=2     \
    actor_rollout_ref.rollout.max_num_batched_tokens=65536     \
    actor_rollout_ref.rollout.name=vllm     \
    actor_rollout_ref.rollout.gpu_memory_utilization=0.6     \
    actor_rollout_ref.rollout.val_kwargs.temperature=0.0     \
    actor_rollout_ref.rollout.val_kwargs.do_sample=False     \
    actor_rollout_ref.rollout.val_kwargs.n=1     \
    actor_rollout_ref.ref.log_prob_micro_batch_size_per_gpu=2     \
    actor_rollout_ref.ref.fsdp_config.param_offload=True     \
    critic.optim.lr=1e-5     \
    critic.model.use_remove_padding=True     \
    critic.model.path=$export_pth     \
    critic.model.enable_gradient_checkpointing=True     \
    critic.ppo_micro_batch_size_per_gpu=2     \
    critic.model.fsdp_config.param_offload=True     \
    critic.model.fsdp_config.optimizer_offload=True     \
    critic.model.fsdp_config.fsdp_size=8     \
    critic.ppo_max_token_len_per_gpu=65536     \
    algorithm.use_process_rewards=False     \
    algorithm.adv_estimator=gae     \
    algorithm.kl_ctrl.kl_coef=0.001     \
    trainer.default_local_dir=./checkpoints/retro_qwen2.5-7b-instruct-1M_10_test/ppo_retro_test190_t0_chembl    \
    trainer.critic_warmup=3     \
    trainer.logger=['console','wandb']     \
    trainer.project_name=retro_qwen2.5-7b-instruct-1M_10_test     \
    trainer.experiment_name=ppo_retro_test190_t0_chembl     \
    trainer.n_gpus_per_node=8     \
    trainer.nnodes=1     \
    trainer.save_freq=10     \
    trainer.test_freq=10     \
    trainer.total_epochs=0     \
    trainer.val_before_train=True     \
    tool.max_turns=100     \
    tool.topk=10     \
    tool.shuffle=False     \
    tool.maxstep=30     \
    tool.force_noloop=True     \
    tool.use_batch_tool_calls=False     \
    tool.env='retro_noback_V4' $@    > ./logs/retro_qwen2.5-7b-instruct-1M_10_test_t0_chembl.log     \
    2> ./logs/retro_qwen2.5-7b-instruct-1M_10_test_t0_chembl_error.log