DocsICP FilingConsole
官方文档
首页

Checkpoint management

更新时间：
复制 MD 格式
产品详情
我的收藏

Manage checkpoints generated by fine-tune jobs: list, export, query validation results, and Checkpoint object reference.

Prerequisites

List checkpoints

List all checkpoints and their status, model information, etc. for a specified fine-tune job.

Request interface

GET https://dashscope.aliyuncs.com/api/v1/fine-tunes/{job_id}/checkpoints

Input parameters

Parameter

Location

Type

Required

Description

Example

job_id

Path

string

Yes

Fine-tune job ID. You can obtain it from the Create fine-tune job or List fine-tune jobs API.

ft-202511111122-xxxx

Response parameters

Parameter

Type

Description

Example

request_id

string

Unique identifier for the request.

aa4b0229-b1db-9afa-bb6e-xxxxxxxxx

output

array[object]

Array of checkpoints, sorted by step in descending order.

For speech synthesis models (such as CosyVoice), the step is derived from the combination of LM epoch and FM epoch. Checkpoints are sorted by the product of LM epoch × FM epoch in descending order (a higher product indicates more thorough fine-tuning on both sides).

-

output[].checkpoint_id

string

Unique checkpoint identifier. Format: {job_id}:checkpoint-{step}. The step format varies by model type; see the step field for details.

ft-202605271743-dd2a:checkpoint-00040004

output[].full_name

string

Same as checkpoint_id.

ft-202605271743-dd2a:checkpoint-00040004

output[].checkpoint

string

Checkpoint name. Format: checkpoint-{step}.

For speech synthesis models (such as CosyVoice), the step is a combination of LM epoch and FM epoch. For example, checkpoint-00040004 means LM epoch 4 + FM epoch 4.

checkpoint-00040004

output[].job_id

string

The fine-tune job ID that this checkpoint belongs to.

ft-202605271743-dd2a

output[].step

integer

Training step number for this checkpoint.

For speech synthesis models (such as CosyVoice), the step is calculated as LM_epoch × 10000 + FM_epoch. For example, 40004 means LM epoch 4 + FM epoch 4.

40004

output[].status

string

Status of this checkpoint. Common values: SUCCEEDED (ready for deployment), PENDING (not yet ready).

SUCCEEDED

output[].model_name

string

Model ID for this checkpoint. Can be used as the model_name parameter in the Deploy Model API. Only returned when status=SUCCEEDED.

cosyvoice-v3-flash-ft-202605271743-dd2a

output[].model_display_name

string

Model display name (shown in the console). Only returned when status=SUCCEEDED.

ft-202605271743-dd2a

output[].create_time

string

Creation time of this checkpoint in ISO 8601 format.

2026-05-27T18:07:16

output[].expire_time

string

Expiration time of this checkpoint in ISO 8601 format.

2026-06-11T18:07:16

Request example

Replace <your-fine-tune-job-id> in the URL with the job_id value from the Create fine-tune job response.

For Windows CMD, replace ${DASHSCOPE_API_KEY} with %DASHSCOPE_API_KEY%. For PowerShell, use $env:DASHSCOPE_API_KEY
curl --location 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<your-fine-tune-job-id>/checkpoints' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json'

Response example

{
    "request_id": "aa4b0229-b1db-9afa-bb6e-3b3e9ee1489b",
    "output": [
        {
            "create_time": "2026-05-27T18:07:16",
            "full_name": "ft-202605271743-dd2a:checkpoint-00040004",
            "job_id": "ft-202605271743-dd2a",
            "checkpoint_id": "ft-202605271743-dd2a:checkpoint-00040004",
            "checkpoint": "checkpoint-00040004",
            "model_name": "cosyvoice-v3-flash-ft-202605271743-dd2a",
            "model_display_name": "ft-202605271743-dd2a",
            "status": "SUCCEEDED",
            "expire_time": "2026-06-11T18:07:16",
            "step": 40004
        },
        {
            "create_time": "2026-05-27T18:07:16",
            "full_name": "ft-202605271743-dd2a:checkpoint-00030004",
            "job_id": "ft-202605271743-dd2a",
            "checkpoint_id": "ft-202605271743-dd2a:checkpoint-00030004",
            "checkpoint": "checkpoint-00030004",
            "status": "PENDING",
            "expire_time": "2026-06-11T18:07:16",
            "step": 30004
        }
    ]
}

Export Checkpoint

Export a specified checkpoint as a deployable model. After successful export, call the corresponding Deploy API based on the model type (different model types use different deploy APIs).

Request interface

GET https://dashscope.aliyuncs.com/api/v1/fine-tunes/{job_id}/export/{checkpoint}?model_name={model_name}

Input parameters

Parameter

Location

Type

Required

Description

Example

job_id

Path

string

Yes

Fine-tune job ID. You can obtain it from the Create fine-tune job or List fine-tune jobs API.

ft-202511111122-xxxx

checkpoint

Path

string

Yes

Checkpoint name. You can obtain it from the List Checkpoints API.

checkpoint-160

model_name

Query parameter

string

Yes

The exported model name displayed in the console.

This name must be globally unique. We recommend using letters, digits, underscores (_), and hyphens (-).

Note: This parameter is only for console display. The actual exported model name is the output[].model_name value from the List Checkpoints response.

wan2.5-i2v-preview-ft-202511111122-xxxx

Response parameters

Parameter

Type

Description

Example

request_id

string

Unique identifier for the request.

0eb05b0c-02ba-414a-9d0c-xxxxxxxxx

output

boolean

Whether the export request was submitted successfully.

  • true: The export request was submitted successfully.

  • false: The export request failed. Please retry.

true

Request example

  • <your-fine-tune-job-id>: Replace with the job_id value from the Create fine-tune job response.

  • <checkpoint-to-export>: Replace with the checkpoint value, e.g., “checkpoint-160”.

  • <export-model-display-name>: Replace with your custom model name (for console display only).

curl --location 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<your-fine-tune-job-id>/export/<checkpoint-to-export>?model_name=<export-model-display-name>' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY"

Response example

{
    "request_id": "0817d1ed-b6b6-4383-9650-xxxxx",
    "output": true
}

List checkpoint validations

Note

This API is not supported for text generation models.

Retrieve the list of checkpoints that successfully generated preview videos or images via the validation set. Checkpoints that fail validation are not listed.

Usage restriction: This API must be called after model fine-tuning is completed; otherwise, an empty list is returned.

Request interface

GET https://dashscope.aliyuncs.com/api/v1/fine-tunes/{job_id}/validation-results

Input parameters

Parameter

Location

Type

Required

Description

Example

job_id

Path

string

Yes

Fine-tune job ID. You can obtain it from the Create fine-tune job or List fine-tune jobs API.

ft-202511111122-xxxx

Response parameters

Parameter

Type

Description

Example

request_id

string

Unique identifier for the request.

0eb05b0c-02ba-414a-9d0c-xxxxxxxxx

output

array[string]

List of checkpoints.

-

output[].checkpoint

string

Checkpoint name.

checkpoint-160

Request example

Replace <your-fine-tune-job-id> in the URL with the job_id value from the Create fine-tune job response.

curl --location 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<your-fine-tune-job-id>/validation-results' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header 'Content-Type: application/json'

Response example

{
    "request_id": "da1310f5-5a21-4e29-99d4-xxxxxx",
    "output": [
        {
            "checkpoint": "checkpoint-160"
        },
        {
            "checkpoint": "checkpoint-20"
        },
        {
            "checkpoint": "checkpoint-40"
        },
        {
            "checkpoint": "checkpoint-60"
        }
    ]
}

Query checkpoint validation results

Note

This API is not supported for text generation models.

Query the validation result details for a specified checkpoint (e.g., "checkpoint-160") by checkpoint, including the generated preview videos or images.

Request interface

GET https://dashscope.aliyuncs.com/api/v1/fine-tunes/{job_id}/validation-details/{checkpoint}?page_no=1&page_size=10

Input parameters

Parameter

Location

Type

Required

Description

Example

job_id

Path

string

Yes

Fine-tune job ID. You can obtain it from the Create fine-tune job or List fine-tune jobs API.

ft-202511111122-xxxx

checkpoint

Path

string

Yes

Checkpoint name. You can obtain it from the List Checkpoints or List Checkpoint Validations API.

checkpoint-160

page_no

Query parameter

integer

No

Page number. Defaults to 1.

1

page_size

Query parameter

integer

No

Page size. Defaults to 10.

10

Response parameters

Parameter

Type

Description

Example

request_id

string

Unique identifier for the request.

375b3ad0-d3fa-451f-b629-xxxxxxx

output

object

Output result.

-

output.page_no

integer

Page number.

1

output.page_size

integer

Page size.

10

output.total

integer

Total number of validation items.

1

output.list

array[object]

List of validation items.

-

output.list[].video_path

string

Video generated by the checkpoint.

The video_path URL is valid for 24 hours. Please download the video promptly.

https://finetune-swap-wulanchabu.oss-cn-wulanchabu.aliyuncs.com/xxx.mp4?Expires=xxxx

output.list[].prompt

string

Prompt for the validation data. Obtained from the dataset annotation file data.jsonl.

The video begins with a scene of a young man sitting in a cafe...

output.list[].first_frame_path

string

Image URL for validation. The system reads the image from the dataset and generates a public URL.

https://finetune-swap-wulanchabu.oss-cn-wulanchabu.aliyuncs.com/xxx.jpeg

Request example

  • <your-fine-tune-job-id>: Replace with the job_id value from the Create fine-tune job response.

  • <selected-checkpoint>: Replace with the selected checkpoint name, e.g., “checkpoint-160”.

curl --location 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<your-fine-tune-job-id>/validation-details/<selected-checkpoint>?page_no=1&page_size=10' \
--header "Authorization: Bearer $DASHSCOPE_API_KEY"

Response example

The video_path URL is valid for 24 hours. Please download the video promptly.
{
    "request_id": "375b3ad0-d3fa-451f-b629-xxxxxxx",
    "output": {
        "page_no": 1,
        "page_size": 10,
        "total": 1,
        "list": [
            {
                "video_path": "https://finetune-swap-wulanchabu.oss-cn-wulanchabu.aliyuncs.com/xxx.mp4?Expires=xxxx",
                "prompt": "The video begins with a young man sitting in a cafe. He wears a beige polo shirt with a focused, slightly contemplative expression, fingers lightly resting on his chin. In front of him is a steaming cup of coffee, with a wooden striped wall and a decorative sign in the background. Then the money rain effect begins, with countless oversized dollar bills pouring down like a storm, densely falling around him. As bills continue to fall, he stretches his arms upward, tilts his head back, and looks delighted, completely immersed in this wild money rain.",
                "first_frame_path": "https://finetune-swap-wulanchabu.oss-cn-wulanchabu.aliyuncs.com/xxx.jpeg"
            }
        ]
    }
}

Checkpoint object

The Checkpoint API is currently only available in the China (Beijing) region. If you are using other regions, manage checkpoints through the Model Studio console of that region.

The Checkpoint object represents an intermediate model state saved during fine-tuning. It is returned by the List Checkpoints API.

Field description

The following fields are from the List Checkpoints API response.

Parameter

Type

Description

request_id

String

Request ID for this call.

output

Array

Array of checkpoints, sorted by step in descending order.

For speech synthesis models (such as CosyVoice), the step is derived from the combination of LM epoch and FM epoch. Checkpoints are sorted by the product of LM epoch × FM epoch in descending order (a higher product indicates more thorough fine-tuning on both sides).

output[*].checkpoint_id

String

Unique checkpoint identifier. Format: {job_id}:checkpoint-{step}. The step format varies by model type; see the step field for details.

output[*].full_name

String

Same as checkpoint_id.

output[*].checkpoint

String

Checkpoint name. Format: checkpoint-{step}.

For speech synthesis models (such as CosyVoice), the step is a combination of LM epoch and FM epoch. For example, checkpoint-00040004 means LM epoch 4 + FM epoch 4.

output[*].job_id

String

The fine-tune job ID that this checkpoint belongs to.

output[*].step

Integer

Training step number for this checkpoint.

For speech synthesis models (such as CosyVoice), the step is calculated as LM_epoch × 10000 + FM_epoch. For example, 40004 means LM epoch 4 + FM epoch 4.

output[*].status

String

Status of this checkpoint. Common values: SUCCEEDED (ready for deployment), PENDING (not yet ready).

output[*].model_name

String

Model ID for this checkpoint. Can be used as the model_name parameter in the Deploy Model API. Only returned when status=SUCCEEDED.

output[*].model_display_name

String

Model display name (shown in the console). Only returned when status=SUCCEEDED.

output[*].create_time

String

Creation time of this checkpoint in ISO 8601 format.

output[*].expire_time

String

Expiration time of this checkpoint in ISO 8601 format.
Previous：Query fine-tuning task detailsNext：Model deployment