训练预估算时间

开发中

POST

/estimator/estimated_train_time

请求参数

Body 参数application/json

示例

{
    "model_name_or_path": "GPT-OSS-20B-Thinking",
    "model_size_in_billion": "20.0",
    "gpu_type": "h100",
    "gpu_nums": 8,
    "public_datasets": [
        // "identity"
        // "alpaca_zh_demo"
        "haruhi_train"
    ],
    "private_datasets": [
        // "QA_from_CoVLA_zh"
        // "CBT-train-mini-3"
        // "alpaca_zh_demo1",
        // "alpaca_zh_demo2",
        // "alpaca_zh_demo3"
    ],
    "user_id": "user123",
    "user_key": "userdddd",
    // "user_id": "7419815b8f0b4f2dae8864359ec893db",
    // "user_key": "user-v3H5RMNi",
    "cutoff_len": 1024,
    "per_device_train_batch_size": 1,
    "gradient_accumulation_steps": 1,
    "num_train_epochs": 1.0,
    "stage": "Supervised Fine-Tuning",
    "finetuning_type": "lora",
    "compute_type": "bf16",
    "booster": "fa2",
    "max_samples": 1000000,
    "lora_rank": 8,
    "lora_alpha": 16,
    "max_grad_norm": 1.0,
    "continue_flag": false,
    "continue_step": 0,
    "packing": false,
    "freeze_trainable_layers": 16,
    "freeze_trainable_modules": "all",
    "modalities": ["text"]
}

请求示例代码

Shell

JavaScript

Java

Swift

PHP

Python

HTTP

Objective-C

Ruby

OCaml

Dart

curl --location --request POST '/estimator/estimated_train_time' \
--header 'Content-Type: application/json' \
--data-raw '{
    "model_name_or_path": "GPT-OSS-20B-Thinking",
    "model_size_in_billion": "20.0",
    "gpu_type": "h100",
    "gpu_nums": 8,
    "public_datasets": [
        // "identity"
        // "alpaca_zh_demo"
        "haruhi_train"
    ],
    "private_datasets": [
        // "QA_from_CoVLA_zh"
        // "CBT-train-mini-3"
        // "alpaca_zh_demo1",
        // "alpaca_zh_demo2",
        // "alpaca_zh_demo3"
    ],
    "user_id": "user123",
    "user_key": "userdddd",
    // "user_id": "7419815b8f0b4f2dae8864359ec893db",
    // "user_key": "user-v3H5RMNi",
    "cutoff_len": 1024,
    "per_device_train_batch_size": 1,
    "gradient_accumulation_steps": 1,
    "num_train_epochs": 1.0,
    "stage": "Supervised Fine-Tuning",
    "finetuning_type": "lora",
    "compute_type": "bf16",
    "booster": "fa2",
    "max_samples": 1000000,
    "lora_rank": 8,
    "lora_alpha": 16,
    "max_grad_norm": 1.0,
    "continue_flag": false,
    "continue_step": 0,
    "packing": false,
    "freeze_trainable_layers": 16,
    "freeze_trainable_modules": "all",
    "modalities": ["text"]
}'

返回响应

🟢200成功

application/json

Body

示例

{
    "predict_total_steps": 250,
    "predict_step_time_sec": 0.432,
    "predict_total_time_sec": 108,
    "predict_total_time_human": "0:01:48",
    "predict_total_time_sec_max": 138,
    "predict_total_time_human_max": "0:02:18",
    "predict_packing_enabled": false,
    "predict_packing_speedup_factor": 0.93
}

修改于 2025-09-09 09:11:36

恢复删除项

评估预估算时间