Img Open

提交训练任务

POST

/AIGCImgOpenServ/saas/modelTraining/submitTraining/Job

img

请求参数

Header 参数

X-Request-req-accessKeyId

string

可选

X-Request-req-accessKeySecret

string

可选

Body 参数application/json

taskId

string

必需

任务id，必须且唯一

fileUrlList

array[string]

必需

训练的图片urllist，必须公网可访问,图片张数 10-100

maxEpochs

integer

可选

迭代步数 10-50 默认20

batchSize

integer

可选

每批次送入图片数量 1-4 默认1

learningRate

string

可选

范围 0.0000001 - 0.0001 默认 0.00001；防止精度丢失str传入

maxTokenLength

integer

可选

范围 0,150,225 默认225

clipSkip

integer

1-5 默认1

可选

randomCrop

boolean

可选

pretrainedModel

string

可选

训练基模型，默认v1-5-pruned-emaonly

repeats

integer

可选

范围 1-30 默认10

mixedPrecision

string

可选

"fp16", "bf16"

savePrecision

string

可选

"fp16", "bf16", "float"

numCpuThreadsPerProcess

integer

1-20

可选

seed

integer

可选

cacheLatents

boolean

可选

cacheLatentsToDisk

boolean

可选

lrScheduler

string

可选

lrWarmupSteps

integer

0-100

可选

optimizerType

string

可选

"AdamW", "AdamW8bit", "PagedAdamW8bit", "Lion8bit", "PagedLion8bit", "Lion", "SGDNesterov", "SGDNesterov8bit", "DAdaptation", "DAdaptAdaGrad", "DAdaptAdam", "DAdaptAdan", "DAdaptAdanIP", "DAdaptLion", "DAdaptSGD", "AdaFactor"

resolution

integer

可选

enableBucket

boolean

可选

minBucketReso

integer

可选

maxBucketReso

integer

可选

textEncoderLr

string

可选

unetLr

string

可选

networkDropout

number

0-1

可选

rankDropout

number

0-1

可选

moduleDropout

number

0-1

可选

networkModule

string

可选

networkDim

integer

可选

networkAlpha

integer

可选

networkArgs

string

可选

boolean

可选

flipAug

boolean

可选

colorAug

boolean

可选

gradientAccumulationSteps

integer

可选

noiseOffset

number

0-1

可选

示例

{
    "file_url_list": [],
    "max_epochs": 20,
    "batch_size": 2,
    "learning_rate": "0.0001",
    "max_token_length": 225,
    "clip_skip": 2,
    "random_crop": 0,
    "repeats": 10,
    "mixedPrecision": "bf16",
    "savePrecision": "bf16",
    "numCpuThreadsPerProcess": 4,
    "seed": 1234,
    "cacheLatents": true,
    "cacheLatentsToDisk": false,
    "lrScheduler": "constant",
    "lrWarmupSteps": 0,
    "optimizerType": "Lion",
    "resolution": 384,
    "enableBucket": true,
    "minBucketReso": 256,
    "maxBucketReso": 2048,
    "textEncoderLr": "5e-05",
    "unetLr": "0.0001",
    "networkDropout": 0,
    "rankDropout": 0,
    "moduleDropout": 0,
    "networkModule": "networks.lora",
    "networkDim": 128,
    "networkAlpha": 128,
    "networkArgs": "",
    "v2": false,
    "flipAug": false,
    "colorAug": false,
    "noiseOffset": 0.05,
}

示例代码

Shell

JavaScript

Java

Swift

PHP

Python

HTTP

Objective-C

Ruby

OCaml

Dart

curl --location --request POST 'https://api.zhiqite.com/AIGCImgOpenServ/saas/modelTraining/submitTraining/Job' \
--header 'X-Request-req-accessKeyId;' \
--header 'X-Request-req-accessKeySecret;' \
--header 'Content-Type: application/json' \
--data-raw '{
    "file_url_list": [],
    "max_epochs": 20,
    "batch_size": 2,
    "learning_rate": "0.0001",
    "max_token_length": 225,
    "clip_skip": 2,
    "random_crop": 0,
    "repeats": 10,
    "mixedPrecision": "bf16",
    "savePrecision": "bf16",
    "numCpuThreadsPerProcess": 4,
    "seed": 1234,
    "cacheLatents": true,
    "cacheLatentsToDisk": false,
    "lrScheduler": "constant",
    "lrWarmupSteps": 0,
    "optimizerType": "Lion",
    "resolution": 384,
    "enableBucket": true,
    "minBucketReso": 256,
    "maxBucketReso": 2048,
    "textEncoderLr": "5e-05",
    "unetLr": "0.0001",
    "networkDropout": 0,
    "rankDropout": 0,
    "moduleDropout": 0,
    "networkModule": "networks.lora",
    "networkDim": 128,
    "networkAlpha": 128,
    "networkArgs": "",
    "v2": false,
    "flipAug": false,
    "colorAug": false,
    "noiseOffset": 0.05,
}'

返回响应

🟢200成功

application/json

Body

code

string

200 成功

必需

msg

string

必需

示例

{
    "code": "string",
    "msg": "string"
}

修改于 2024-01-15 10:12:39

绘图工具任务查询

训练结果回调