Create Rl Training Job

Create a new RL (Reinforcement Learning) training job.

POST /v1/preview/training-jobs

Create Rl Training Job

cURL

curl --request POST \
  --url https://api.training.wandb.ai/v1/preview/training-jobs \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "model_id": "3c90c3cc-0d44-4b50-8888-8dd25736052a",
  "trajectory_groups": [
    {
      "trajectories": [
        {
          "messages_and_choices": [],
          "tools": [
            {
              "function": {
                "name": "<string>",
                "description": "<string>",
                "parameters": {},
                "strict": true
              },
              "type": "<string>"
            }
          ],
          "additional_histories": [],
          "reward": 0,
          "initial_policy_version": 123,
          "final_policy_version": 123,
          "metrics": {},
          "metadata": {},
          "logs": [],
          "start_time": "2023-11-07T05:31:56Z"
        }
      ]
    }
  ],
  "experimental_config": {
    "advantage_balance": 123,
    "allow_training_without_logprobs": true,
    "epsilon": 123,
    "epsilon_high": 123,
    "kimi_k2_tau": 123,
    "kl_penalty_coef": 123,
    "kl_penalty_reference_step": 123,
    "kl_penalty_step_lag": 123,
    "kl_ref_adapter_path": "<string>",
    "learning_rate": 123,
    "logprob_calculation_chunk_size": 123,
    "mask_prob_ratio": true,
    "max_negative_advantage_importance_sampling_weight": 123,
    "normalize_advantages": true,
    "num_trajectories_learning_rate_multiplier_power": 123,
    "packed_sequence_length": 123,
    "plot_tensors": true,
    "ppo": true,
    "precalculate_logprobs": true,
    "scale_learning_rate_by_reward_std_dev": true,
    "scale_rewards": true,
    "truncated_importance_sampling": 123
  }
}
'

import requests

url = "https://api.training.wandb.ai/v1/preview/training-jobs"

payload = {
    "model_id": "3c90c3cc-0d44-4b50-8888-8dd25736052a",
    "trajectory_groups": [{ "trajectories": [
                {
                    "messages_and_choices": [],
                    "tools": [
                        {
                            "function": {
                                "name": "<string>",
                                "description": "<string>",
                                "parameters": {},
                                "strict": True
                            },
                            "type": "<string>"
                        }
                    ],
                    "additional_histories": [],
                    "reward": 0,
                    "initial_policy_version": 123,
                    "final_policy_version": 123,
                    "metrics": {},
                    "metadata": {},
                    "logs": [],
                    "start_time": "2023-11-07T05:31:56Z"
                }
            ] }],
    "experimental_config": {
        "advantage_balance": 123,
        "allow_training_without_logprobs": True,
        "epsilon": 123,
        "epsilon_high": 123,
        "kimi_k2_tau": 123,
        "kl_penalty_coef": 123,
        "kl_penalty_reference_step": 123,
        "kl_penalty_step_lag": 123,
        "kl_ref_adapter_path": "<string>",
        "learning_rate": 123,
        "logprob_calculation_chunk_size": 123,
        "mask_prob_ratio": True,
        "max_negative_advantage_importance_sampling_weight": 123,
        "normalize_advantages": True,
        "num_trajectories_learning_rate_multiplier_power": 123,
        "packed_sequence_length": 123,
        "plot_tensors": True,
        "ppo": True,
        "precalculate_logprobs": True,
        "scale_learning_rate_by_reward_std_dev": True,
        "scale_rewards": True,
        "truncated_importance_sampling": 123
    }
}
headers = {
    "Authorization": "Bearer <token>",
    "Content-Type": "application/json"
}

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

print(response.text)

const options = {
  method: 'POST',
  headers: {Authorization: 'Bearer <token>', 'Content-Type': 'application/json'},
  body: JSON.stringify({
    model_id: '3c90c3cc-0d44-4b50-8888-8dd25736052a',
    trajectory_groups: [
      {
        trajectories: [
          {
            messages_and_choices: [],
            tools: [
              {
                function: {name: '<string>', description: '<string>', parameters: {}, strict: true},
                type: '<string>'
              }
            ],
            additional_histories: [],
            reward: 0,
            initial_policy_version: 123,
            final_policy_version: 123,
            metrics: {},
            metadata: {},
            logs: [],
            start_time: '2023-11-07T05:31:56Z'
          }
        ]
      }
    ],
    experimental_config: {
      advantage_balance: 123,
      allow_training_without_logprobs: true,
      epsilon: 123,
      epsilon_high: 123,
      kimi_k2_tau: 123,
      kl_penalty_coef: 123,
      kl_penalty_reference_step: 123,
      kl_penalty_step_lag: 123,
      kl_ref_adapter_path: '<string>',
      learning_rate: 123,
      logprob_calculation_chunk_size: 123,
      mask_prob_ratio: true,
      max_negative_advantage_importance_sampling_weight: 123,
      normalize_advantages: true,
      num_trajectories_learning_rate_multiplier_power: 123,
      packed_sequence_length: 123,
      plot_tensors: true,
      ppo: true,
      precalculate_logprobs: true,
      scale_learning_rate_by_reward_std_dev: true,
      scale_rewards: true,
      truncated_importance_sampling: 123
    }
  })
};

fetch('https://api.training.wandb.ai/v1/preview/training-jobs', options)
  .then(res => res.json())
  .then(res => console.log(res))
  .catch(err => console.error(err));

<?php

$curl = curl_init();

curl_setopt_array($curl, [
  CURLOPT_URL => "https://api.training.wandb.ai/v1/preview/training-jobs",
  CURLOPT_RETURNTRANSFER => true,
  CURLOPT_ENCODING => "",
  CURLOPT_MAXREDIRS => 10,
  CURLOPT_TIMEOUT => 30,
  CURLOPT_HTTP_VERSION => CURL_HTTP_VERSION_1_1,
  CURLOPT_CUSTOMREQUEST => "POST",
  CURLOPT_POSTFIELDS => json_encode([
    'model_id' => '3c90c3cc-0d44-4b50-8888-8dd25736052a',
    'trajectory_groups' => [
        [
                'trajectories' => [
                                [
                                                                'messages_and_choices' => [
                                                                                                                                
                                                                ],
                                                                'tools' => [
                                                                                                                                [
                                                                                                                                                                                                                                                                'function' => [
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                'name' => '<string>',
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                'description' => '<string>',
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                'parameters' => [
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                ],
                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                'strict' => true
                                                                                                                                                                                                                                                                ],
                                                                                                                                                                                                                                                                'type' => '<string>'
                                                                                                                                ]
                                                                ],
                                                                'additional_histories' => [
                                                                                                                                
                                                                ],
                                                                'reward' => 0,
                                                                'initial_policy_version' => 123,
                                                                'final_policy_version' => 123,
                                                                'metrics' => [
                                                                                                                                
                                                                ],
                                                                'metadata' => [
                                                                                                                                
                                                                ],
                                                                'logs' => [
                                                                                                                                
                                                                ],
                                                                'start_time' => '2023-11-07T05:31:56Z'
                                ]
                ]
        ]
    ],
    'experimental_config' => [
        'advantage_balance' => 123,
        'allow_training_without_logprobs' => true,
        'epsilon' => 123,
        'epsilon_high' => 123,
        'kimi_k2_tau' => 123,
        'kl_penalty_coef' => 123,
        'kl_penalty_reference_step' => 123,
        'kl_penalty_step_lag' => 123,
        'kl_ref_adapter_path' => '<string>',
        'learning_rate' => 123,
        'logprob_calculation_chunk_size' => 123,
        'mask_prob_ratio' => true,
        'max_negative_advantage_importance_sampling_weight' => 123,
        'normalize_advantages' => true,
        'num_trajectories_learning_rate_multiplier_power' => 123,
        'packed_sequence_length' => 123,
        'plot_tensors' => true,
        'ppo' => true,
        'precalculate_logprobs' => true,
        'scale_learning_rate_by_reward_std_dev' => true,
        'scale_rewards' => true,
        'truncated_importance_sampling' => 123
    ]
  ]),
  CURLOPT_HTTPHEADER => [
    "Authorization: Bearer <token>",
    "Content-Type: application/json"
  ],
]);

$response = curl_exec($curl);
$err = curl_error($curl);

curl_close($curl);

if ($err) {
  echo "cURL Error #:" . $err;
} else {
  echo $response;
}

package main

import (
	"fmt"
	"strings"
	"net/http"
	"io"
)

func main() {

	url := "https://api.training.wandb.ai/v1/preview/training-jobs"

	payload := strings.NewReader("{\n  \"model_id\": \"3c90c3cc-0d44-4b50-8888-8dd25736052a\",\n  \"trajectory_groups\": [\n    {\n      \"trajectories\": [\n        {\n          \"messages_and_choices\": [],\n          \"tools\": [\n            {\n              \"function\": {\n                \"name\": \"<string>\",\n                \"description\": \"<string>\",\n                \"parameters\": {},\n                \"strict\": true\n              },\n              \"type\": \"<string>\"\n            }\n          ],\n          \"additional_histories\": [],\n          \"reward\": 0,\n          \"initial_policy_version\": 123,\n          \"final_policy_version\": 123,\n          \"metrics\": {},\n          \"metadata\": {},\n          \"logs\": [],\n          \"start_time\": \"2023-11-07T05:31:56Z\"\n        }\n      ]\n    }\n  ],\n  \"experimental_config\": {\n    \"advantage_balance\": 123,\n    \"allow_training_without_logprobs\": true,\n    \"epsilon\": 123,\n    \"epsilon_high\": 123,\n    \"kimi_k2_tau\": 123,\n    \"kl_penalty_coef\": 123,\n    \"kl_penalty_reference_step\": 123,\n    \"kl_penalty_step_lag\": 123,\n    \"kl_ref_adapter_path\": \"<string>\",\n    \"learning_rate\": 123,\n    \"logprob_calculation_chunk_size\": 123,\n    \"mask_prob_ratio\": true,\n    \"max_negative_advantage_importance_sampling_weight\": 123,\n    \"normalize_advantages\": true,\n    \"num_trajectories_learning_rate_multiplier_power\": 123,\n    \"packed_sequence_length\": 123,\n    \"plot_tensors\": true,\n    \"ppo\": true,\n    \"precalculate_logprobs\": true,\n    \"scale_learning_rate_by_reward_std_dev\": true,\n    \"scale_rewards\": true,\n    \"truncated_importance_sampling\": 123\n  }\n}")

	req, _ := http.NewRequest("POST", url, payload)

	req.Header.Add("Authorization", "Bearer <token>")
	req.Header.Add("Content-Type", "application/json")

	res, _ := http.DefaultClient.Do(req)

	defer res.Body.Close()
	body, _ := io.ReadAll(res.Body)

	fmt.Println(string(body))

}

HttpResponse<String> response = Unirest.post("https://api.training.wandb.ai/v1/preview/training-jobs")
  .header("Authorization", "Bearer <token>")
  .header("Content-Type", "application/json")
  .body("{\n  \"model_id\": \"3c90c3cc-0d44-4b50-8888-8dd25736052a\",\n  \"trajectory_groups\": [\n    {\n      \"trajectories\": [\n        {\n          \"messages_and_choices\": [],\n          \"tools\": [\n            {\n              \"function\": {\n                \"name\": \"<string>\",\n                \"description\": \"<string>\",\n                \"parameters\": {},\n                \"strict\": true\n              },\n              \"type\": \"<string>\"\n            }\n          ],\n          \"additional_histories\": [],\n          \"reward\": 0,\n          \"initial_policy_version\": 123,\n          \"final_policy_version\": 123,\n          \"metrics\": {},\n          \"metadata\": {},\n          \"logs\": [],\n          \"start_time\": \"2023-11-07T05:31:56Z\"\n        }\n      ]\n    }\n  ],\n  \"experimental_config\": {\n    \"advantage_balance\": 123,\n    \"allow_training_without_logprobs\": true,\n    \"epsilon\": 123,\n    \"epsilon_high\": 123,\n    \"kimi_k2_tau\": 123,\n    \"kl_penalty_coef\": 123,\n    \"kl_penalty_reference_step\": 123,\n    \"kl_penalty_step_lag\": 123,\n    \"kl_ref_adapter_path\": \"<string>\",\n    \"learning_rate\": 123,\n    \"logprob_calculation_chunk_size\": 123,\n    \"mask_prob_ratio\": true,\n    \"max_negative_advantage_importance_sampling_weight\": 123,\n    \"normalize_advantages\": true,\n    \"num_trajectories_learning_rate_multiplier_power\": 123,\n    \"packed_sequence_length\": 123,\n    \"plot_tensors\": true,\n    \"ppo\": true,\n    \"precalculate_logprobs\": true,\n    \"scale_learning_rate_by_reward_std_dev\": true,\n    \"scale_rewards\": true,\n    \"truncated_importance_sampling\": 123\n  }\n}")
  .asString();

require 'uri'
require 'net/http'

url = URI("https://api.training.wandb.ai/v1/preview/training-jobs")

http = Net::HTTP.new(url.host, url.port)
http.use_ssl = true

request = Net::HTTP::Post.new(url)
request["Authorization"] = 'Bearer <token>'
request["Content-Type"] = 'application/json'
request.body = "{\n  \"model_id\": \"3c90c3cc-0d44-4b50-8888-8dd25736052a\",\n  \"trajectory_groups\": [\n    {\n      \"trajectories\": [\n        {\n          \"messages_and_choices\": [],\n          \"tools\": [\n            {\n              \"function\": {\n                \"name\": \"<string>\",\n                \"description\": \"<string>\",\n                \"parameters\": {},\n                \"strict\": true\n              },\n              \"type\": \"<string>\"\n            }\n          ],\n          \"additional_histories\": [],\n          \"reward\": 0,\n          \"initial_policy_version\": 123,\n          \"final_policy_version\": 123,\n          \"metrics\": {},\n          \"metadata\": {},\n          \"logs\": [],\n          \"start_time\": \"2023-11-07T05:31:56Z\"\n        }\n      ]\n    }\n  ],\n  \"experimental_config\": {\n    \"advantage_balance\": 123,\n    \"allow_training_without_logprobs\": true,\n    \"epsilon\": 123,\n    \"epsilon_high\": 123,\n    \"kimi_k2_tau\": 123,\n    \"kl_penalty_coef\": 123,\n    \"kl_penalty_reference_step\": 123,\n    \"kl_penalty_step_lag\": 123,\n    \"kl_ref_adapter_path\": \"<string>\",\n    \"learning_rate\": 123,\n    \"logprob_calculation_chunk_size\": 123,\n    \"mask_prob_ratio\": true,\n    \"max_negative_advantage_importance_sampling_weight\": 123,\n    \"normalize_advantages\": true,\n    \"num_trajectories_learning_rate_multiplier_power\": 123,\n    \"packed_sequence_length\": 123,\n    \"plot_tensors\": true,\n    \"ppo\": true,\n    \"precalculate_logprobs\": true,\n    \"scale_learning_rate_by_reward_std_dev\": true,\n    \"scale_rewards\": true,\n    \"truncated_importance_sampling\": 123\n  }\n}"

response = http.request(request)
puts response.read_body

200

{
  "id": "3c90c3cc-0d44-4b50-8888-8dd25736052a"
}

{
  "detail": [
    {
      "loc": [
        "<string>"
      ],
      "msg": "<string>",
      "type": "<string>"
    }
  ]
}

Authorizations

Authorization · string · header · required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json

Schema for creating a new TrainingJob.

model_id · string<uuid> · required

trajectory_groups · TrajectoryGroup · object[] · required

Show child attributes

experimental_config · ExperimentalTrainingConfig · object | null

Schema for experimental training config.

Show child attributes

Response

200

application/json

Successful Response

Schema for TrainingJob response.

id · string<uuid> · required