Trinity-RFT/examples/opmd_gsm8k/opmd_gsm8k.yaml at 0d0276bbf313baadba2d1cf6ed4ff2a45a5a61af · chenyushuo/Trinity-RFT · 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
data:
  total_epochs: 1
  batch_size: 96
  dataset_path: '{path to datasets}/gsm8k'
  default_workflow_type: 'math_workflow'
  format_config:
    prompt_key: 'question'
    response_key: 'answer'
model:
  model_path: '{path to models}/Qwen2.5-1.5B-Inst'
  max_prompt_tokens: 256
  max_response_tokens: 1024
  checkpoint_path: '{path to checkpoints}/test-opmd-gsm8k/qwen2.5-1.5B-gsm8k-opmd-kl_0.001-entropy_0-tau_4-beta1_0.0-beta2_0.95-lr_2e-6-sync10'
cluster:
  node_num: 1
  gpu_per_node: 8
buffer:
  max_retry_times: 3
  max_retry_interval: 1
  train_dataset:
    name: gsm8k_buffer
    storage_type: queue
    path: 'sqlite:///gsm8k_opmd.db'
explorer:
  engine_type: vllm_async
  engine_num: 2
  runner_num: 32
  tensor_parallel_size: 1
  enable_prefix_caching: false
  enforce_eager: true
  dtype: bfloat16
  temperature: 1.0
  seed: 42
  logprobs: 0
  repeat_times: 8
  use_ray: false
  backend: 'nccl'
  max_pending_requests: 32
  max_waiting_steps: 4
synchronizer:
  sync_method: 'nccl'
  sync_iteration_interval: 10
  sync_timeout: 1200
trainer:
  trainer_type: 'verl'
  algorithm_type: opmd
  trainer_config_path: 'examples/opmd_gsm8k/train_opmd_gsm8k.yaml'
  sft_warmup_iteration: 0
  save_interval: 100
monitor:
  cache_root_dir: ""
  project: "Trinity-RFT-gsm8k-test-opmd"
  name: "qwen2.5-1.5B-gsm8k-opmd-kl_0.001-entropy_0-tau_4-beta1_0.0-beta2_0.95-lr_2e-6-sync10"