创建分布式训练任务

curl -X 'POST' \
  'https://api.alayanew.com/v1/subscription/training/create' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -H 'Authorization: Bearer [YOUR_API_KEY]' \
  -d '{
    "name": "my training task",
    "createType": "DIRECT",
    "desc": "my training task",
    "priority": 3,
    "trainingType": "PRE_TRAINING",
    "imageType": "general",
    "trainingFramework": "PyTorch",
    "productCode": "PRD-QTT",
    "workerCount": 1,
    "image": "harbor.unite.zetyun.cn/alayanew-87cbf7ca-f126-4194-b137-8fde5adcd09b/fq-torch-1120:latest",
    "storageConfigs": [
      {
        "storageId": "72a2a885-e45e-4c79-aaf3-e1fa05abdb92",
        "storageType": "nas-capacity",
        "fileDirectory": "nas123",
        "mountPath": "/root/nas/",
        "onlyRead": true
      }
    ],
    "env": {
      "CUDA_VISIBLE_DEVICES": "0",
      "CUDA_VISIBLE_MASTER": "yu"
    },
    "enableAutoRetry": true,
    "maxRetryCount": 3,
    "enableTimeoutCancel": true,
    "timeoutHours": 1,
    "startCommand": "cat > /tmp/mnist.py << 'EOF'\nimport torch\nimport torch.distributed as dist\nimport torch.nn as nn\nimport torch.optim as optim\nfrom torch.nn.parallel import DistributedDataParallel as DDP\nimport os\n\ndef setup_device():\n    if torch.cuda.is_available():\n        local_rank = int(os.environ.get(\"LOCAL_RANK\", 0))\n        device = torch.device(f\"cuda:{local_rank}\")\n        torch.cuda.set_device(device)\n        backend = \"nccl\"\n    else:\n        device = torch.device(\"cpu\")\n        backend = \"gloo\"\n    return device, backend\n\ndef main():\n    device, backend = setup_device()\n    dist.init_process_group(backend=backend)\n    rank = dist.get_rank()\n    world_size = dist.get_world_size()\n\n    model = nn.Linear(10, 10).to(device)\n\n    if backend == \"nccl\":\n        model = DDP(model, device_ids=[device.index])\n    else:\n        model = DDP(model)\n\n    inputs = torch.randn(20, 10).to(device)\n    labels = torch.randn(20, 10).to(device)\n\n    optimizer = optim.SGD(model.parameters(), lr=0.01)\n    loss_fn = nn.MSELoss()\n\n    for epoch in range(10):\n        optimizer.zero_grad()\n        outputs = model(inputs)\n        loss = loss_fn(outputs, labels)\n        loss.backward()\n        optimizer.step()\n\n        if rank == 0:\n            print(f\"Epoch {epoch}, Loss: {loss.item():.4f}, World Size: {world_size}\")\n\n    dist.destroy_process_group()\n\nif __name__ == \"__main__\":\n    main()\nEOF\npython3 /tmp/mnist.py"
  }'

import requests

url = "https://api.alayanew.com/v1/subscription/training/create"
headers = {
    "accept": "application/json",
    "Content-Type": "application/json",
    "Authorization": "Bearer [YOUR_API_KEY]"
}
payload = {
    "name": "llama3-8b-sft",
    "createType": "DIRECT",
    "desc": "Llama3-8B 指令微调",
    "priority": 3,
    "trainingType": "pre_training",
    "imageType": "general",
    "trainingFramwork": "pytorch",
    "productCode": "PRD-TRAIN-1",
    "workerCount": 4,
    "image": "harbor.zetyun.cn/anc-public/general/pytorch:2.3.1-gpu",
    "storageConfigs": [
        {
            "storageId": "72a2a885-e45e-4c79-aaf3-e1fa05abdb92",
            "storageType": "nas-capacity",
            "fileDirectory": "datasets",
            "mountPath": "/root/nas/",
            "onlyRead": False
        }
    ],
    "env": {"CUDA_VISIBLE_DEVICES": "0,1,2,3"},
    "enableAutoRetry": True,
    "maxRetryCount": 3,
    "enableTimeoutCancel": True,
    "timeoutHours": 24,
    "startCommand": "python3 -m torch.distributed.run train.py --data /root/nas/datasets"
}

response = requests.post(url, headers=headers, json=payload)
response.raise_for_status()
print(response.json())

const payload = {
  name: 'llama3-8b-sft',
  createType: 'DIRECT',
  desc: 'Llama3-8B 指令微调',
  priority: 3,
  trainingType: 'pre_training',
  imageType: 'general',
  trainingFramwork: 'pytorch',
  productCode: 'PRD-TRAIN-1',
  workerCount: 4,
  image: 'harbor.zetyun.cn/anc-public/general/pytorch:2.3.1-gpu',
  storageConfigs: [
    {
      storageId: '72a2a885-e45e-4c79-aaf3-e1fa05abdb92',
      storageType: 'nas-capacity',
      fileDirectory: 'datasets',
      mountPath: '/root/nas/',
      onlyRead: false
    }
  ],
  env: { CUDA_VISIBLE_DEVICES: '0,1,2,3' },
  enableAutoRetry: true,
  maxRetryCount: 3,
  enableTimeoutCancel: true,
  timeoutHours: 24,
  startCommand: 'python3 -m torch.distributed.run train.py --data /root/nas/datasets'
};

fetch('https://api.alayanew.com/v1/subscription/training/create', {
  method: 'POST',
  headers: {
    'accept': 'application/json',
    'Content-Type': 'application/json',
    'Authorization': 'Bearer [YOUR_API_KEY]'
  },
  body: JSON.stringify(payload)
})
  .then(res => {
    if (!res.ok) {
      throw new Error(`HTTP error! status: ${res.status}`);
    }
    return res.json();
  })
  .then(console.log)
  .catch(console.error);

{
  "status": 200,
  "message": "OK",
  "data": "a1b2c3d4-0000-1111-2222-333344445555"
}

{
  "status": 403,
  "message": "Forbidden",
  "data": {}
}

{
  "status": 500,
  "message": "Internal Server Error",
  "data": {}
}