verl-loras/.ml-job-preset.yml at main · corl-team/verl-loras · 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
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
distributed:
  time_limit: 32h
  flavor: h100-1x
  generate_name: verl
  master:
    job:
      priority: high
      time_limit: 32h0m0s
      image: docker-hosted.artifactory.tcsbank.ru/trs/verl:1
      region: ix-m5-sm11
      work_dir: /workspace
      env:
          DEEPSPEED_ENABLED: true
          TOKENIZERS_PARALLELISM: false
          NCCL_CUMEM_ENABLE: 0
          HF_DATASETS_CACHE: /workspace/hf_cache/
      args:
        - bash
        - run_master.sh
      input:
        - type: files
          src: ./verl
          dst: "/workspace/verl"
          mask: "**/*.py"
        - type: files
          src: ./
          dst: "/workspace/"
          mask: "**/*.sh"

         # model
        - type: s3msk
          src: reasoning/models/Qwen2.5-Math-1.5B
          dst: /from_s3/models/Qwen2.5-Math-1.5B
          bucket: trs-vsiniy
          cache:
            location: cluster
            enable: true
            created_after: "2025-05-04T21:50:01+0000"

        # data
        - type: s3msk
          src: reasoning/datasets/hf/gsm8k
          dst: /from_s3/datasets/
          bucket: trs-vsiniy
          cache:
            location: cluster
            enable: true
            created_after: "2025-04-16T08:50:01+0000"

  workers:
    - job:
        priority: high
        time_limit: 32h
        region: ix-m5-sm11
        replicas: 1
        image: docker-hosted.artifactory.tcsbank.ru/trs/verl:1
        work_dir: /workspace
        env:
            DEEPSPEED_ENABLED: true
            TOKENIZERS_PARALLELISM: false
            NCCL_CUMEM_ENABLE: 0
            HF_DATASETS_CACHE: /workspace/hf_cache/
        args:
          - bash
          - run_worker.sh
        input:
          - type: files
            src: ./verl
            dst: "/workspace/verl"
            mask: "**/*"
          - type: files
            src: ./
            dst: "/workspace/"
            mask: "**/*.sh"

          # model
          - type: s3msk
            src: reasoning/models/Qwen2.5-Math-1.5B
            dst: /from_s3/models/Qwen2.5-Math-1.5B
            bucket: trs-vsiniy
            cache:
              location: cluster
              enable: true
              created_after: "2025-05-04T21:50:01+0000"

          # data
          - type: s3msk
            src: reasoning/datasets/hf/gsm8k
            dst: /from_s3/datasets/
            bucket: trs-vsiniy
            cache:
              location: cluster
              enable: true
              created_after: "2025-04-16T08:50:01+0000"