Fine-tuning operations - Alibaba Cloud Model Studio - Alibaba Cloud Documentation Center

Learn how to tune Qwen models in Alibaba Cloud Model Studio using the API (HTTP) and the command line (shell). Model tuning involves three methods: supervised fine-tuning (SFT), continual pre-training (CPT), and direct preference optimization (DPO).

Prerequisites

Review Introduction to model fine-tuning to understand its concepts, process, and data requirements.
Activate the service and obtain an API key. For instructions, see Get an API key.
Grant the RAM user (RAM user) the necessary invocation, training, and deployment permissions.

Note

The API supports only token-based billing for training jobs. To use model training units (prepaid or postpaid), create the job in the console.

Tuning file upload

Preparing fine-tuning files

SFT training set

SFT ChatML (Chat Markup Language) format training data supports multi-turn conversations and various role settings.

The OpenAI name and weight parameters are not supported. All assistant outputs will be trained.

# A single line of training data (in JSON format) has the following typical structure when expanded:
{"messages": [
  {"role": "system", "content": "System input 1"}, 
  {"role": "user", "content": "User input 1"}, 
  {"role": "assistant", "content": "Expected model output 1"}, 
  {"role": "user", "content": "User input 2"}, 
  {"role": "assistant", "content": "Expected model output 2"}
  ...
]}

For information about the differences between system, user, and assistant, see Overview. Sample training datasets: SFT-ChatML_format_example.jsonl, SFT-ChatML_format_example.xlsx. XLS and XLSX formats support only single-turn conversations.

All assistant lines in a single training data entry support the "loss_weight" parameter, which sets the relative importance of that line during training. (Range: 0.0 to `1.0`. A larger value indicates higher importance.)

This parameter is available for invitational preview. To use it, contact your account manager.

 {"role": "assistant", "content": "Expected model output 1", "loss_weight": 1.0}, 
 {"role": "assistant", "content": "Expected model output 2", "loss_weight": 0.5}

You can also download a data template from the Model Studio console.

Upload fine-tuning file to Model Studio

OpenAI-compatible Files API

Python

import os
from pathlib import Path
from openai import OpenAI


client = OpenAI(
    # If you have not configured an environment variable, replace the following line with api_key="sk-xxx" and use your Model Studio API key.
    # API keys for the Singapore and China (Beijing) regions are different. Get an API key: https://www.alibabacloud.com/help/en/model-studio/get-api-key
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    # The following is the URL for the Singapore region. If you use a service in the China (Beijing) region, replace the URL with: https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1
    base_url="https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1",
)

# test.jsonl is a local sample file.
file_object = client.files.create(file=Path("test.jsonl"), purpose="fine-tune")

print(file_object.model_dump_json())

Java

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.models.files.*;

import java.nio.file.Path;
import java.nio.file.Paths;

public class Main {
    public static void main(String[] args) {
        // Create a client and use the API key from the environment variable.
        OpenAIClient client = OpenAIOkHttpClient.builder()
                // API keys for the Singapore and China (Beijing) regions are different. Get an API key: https://www.alibabacloud.com/help/en/model-studio/get-api-key
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                // The following is the URL for the Singapore region. If you use a service in the China (Beijing) region, replace the URL with: https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1
                .baseUrl("https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1")
                .build();
        // Set the file path. Modify the path and filename as needed.
        Path filePath = Paths.get("src/main/java/org/example/test.txt");
        // Create file upload parameters.
        FileCreateParams params = FileCreateParams.builder()
                .file(filePath)
                .purpose(FilePurpose.of("fine-tune"))
                .build();

        // Upload the file.
        FileObject fileObject = client.files().create(params);
        System.out.println(fileObject);
    }
}

curl

# ======= Important =======
# API keys for the Singapore and China (Beijing) regions are different. Get an API key: https://www.alibabacloud.com/help/en/model-studio/get-api-key
# The following is the URL for the Singapore region. If you use a service in the China (Beijing) region, replace the URL with: https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/files
# === Delete this comment before running ===

curl -X POST https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/files \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
--form 'file=@"test.jsonl"' \
--form 'purpose="fine-tune"'

Note

Limitations:

The maximum size of a single file is 300 MB.
The total size of all non-deleted files is limited to 5 GB.
You can store a maximum of 100 non-deleted files.
Files are stored indefinitely.

Model fine-tuning

Create a fine-tuning job

HTTP

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" \
--header "Authorization: Bearer ${DASHSCOPE_API_KEY}" \
--header 'Content-Type: application/json' \
--data '{
    "model":"qwen3-8b",
    "training_file_ids":[
        "<your_training_file_id_1>",
        "<your_training_file_id_2>"
    ],
    "hyper_parameters":
    {
        "n_epochs": 3,
        "batch_size": 16,
        "max_length": 8192,
        "learning_rate": "1.6e-5",
        "lr_scheduler_type": "linear",
        "split": 0.9,
        "warmup_ratio": 0.05,
        "eval_steps": 50,
        "data_augmentation": true,
        "augmentation_ratio": "0.1,0.05,0.15",
        "augmentation_types": "dialogue_CN,general_purpose_CN,NLP",
        "save_strategy": "epoch",
        "save_total_limit": 10
    },
    "training_type":"sft"
}'

Parameters

Parameter	Required	Type	Location	Description
training_file_ids	Yes	Array	Body	A list of file IDs for the training set.
validation_file_ids	No	Array	Body	A list of file IDs for the validation set.
model	Yes	String	Body	The ID of the base model for fine-tuning, or the ID of a previously fine-tuned model.
hyper_parameters	No	Map	Body	Hyperparameters for the fine-tuning job. Supported parameters and their default values vary by model. To view the default values, go to the console and select the same model and fine-tuning method. The following parameters are required because they affect the training cost: `n_epochs` (number of epochs), `batch_size` (batch size), and `max_length` (sequence length).
training_type	No	String	Body	Specifies the fine-tuning method. Valid values are: `cpt` `sft` `efficient_sft` `dpo_full` `dpo_lora`
job_name	No	String	Body	Specifies the name of the fine-tuning job.
model_name	No	String	Body	Specifies the name of the fine-tuned model. This is not the model ID, which is generated by the system.

Response

{
    "request_id": "635f7047-003e-4be3-b1db-6f98e239f57b",
    "output":
    {
        "job_id": "ft-202511272033-8ae7",
        "job_name": "ft-202511272033-8ae7",
        "status": "PENDING",
        "finetuned_output": "qwen3-8b-ft-202511272033-8ae7",
        "model": "qwen3-8b",
        "base_model": "qwen3-8b",
        "training_file_ids":
        [
            "9e9ffdfa-c3bf-436e-9613-6f053c66aa6e"
        ],
        "validation_file_ids":
        [],
        "hyper_parameters":
        {
            "n_epochs": 3,
            "batch_size": 16,
            "max_length": 8192,
            "learning_rate": "1.6e-5",
            "lr_scheduler_type": "linear",
            "split": 0.9,
            "warmup_ratio": 0.05,
            "eval_steps": 50,
            "data_augmentation": true,
            "augmentation_ratio": "0.1,0.05,0.15",
            "augmentation_types": "dialogue_CN,general_purpose_CN,NLP",
            "save_strategy": "epoch",
            "save_total_limit": 10
        },
        "training_type": "sft",
        "create_time": "2025-11-27 20:33:15",
        "workspace_id": "llm-8v53etv3hwb8orx1",
        "user_identity": "1654290265984853",
        "modifier": "1654290265984853",
        "creator": "1654290265984853",
        "group": "llm",
        "max_output_cnt": 10
    }
}

Base models (`model`) and training types (`training_type`)

Supported models

Singapore

Text generation

Model name	Model code	SFT full-parameter training (sft)	SFT efficient training (efficient_sft)
Qwen3-14B	qwen3-14b	×	Supported

Visual understanding (Qwen-VL)

Model name	Model code	SFT full-parameter training (sft)	SFT efficient training (efficient_sft)
-	-	-	-

North China 2 (Beijing)

Text generation

Model service	Model code	CPT full-parameter training (cpt)	SFT full-parameter training (sft)	SFT efficient training (efficient_sft)	DPO full-parameter training (dpo_full)	DPO efficient training (dpo_lora)
Qwen3.6-Flash-2026-04-16	qwen3.6-flash-2026-04-16	×	Supported	×	×	×

Qwen3.5-27B	qwen3.5-27b	×	Supported	Supported	×	×
Qwen3.5-9B	qwen3.5-9b	×	Supported	Supported	×	×
Qwen3.5-Flash-2026-02-23	qwen3.5-flash-2026-02-23	×	Supported	×	×	×

Qwen3-32B	qwen3-32b	Supported	Supported	Supported	Supported	Supported
Qwen3-30B-A3B-Instruct-2507	qwen3-30b-a3b-instruct-2507	Supported	Supported	Supported	×	×
Qwen3-14B	qwen3-14b	×	Supported	Supported	Supported	Supported
Qwen3-8B	qwen3-8b	×	Supported	Supported	Supported	Supported
Qwen3-4B-Instruct-2507	qwen3-4b-instruct-2507	Supported	Supported	Supported	Supported	Supported
Qwen3-1.7B	qwen3-1.7b	Supported	Supported	Supported	Supported	Supported
Qwen3-0.6B	qwen3-0.6b	Supported	Supported	Supported	Supported	Supported

Qwen2.5-72B-Instruct	qwen2.5-72b-instruct	Supported	Supported	Supported	Supported	Supported
Qwen2.5-32B-Instruct	qwen2.5-32b-instruct	Supported	Supported	Supported	Supported	Supported
Qwen2.5-14B-Instruct	qwen2.5-14b-instruct	Supported	Supported	Supported	Supported	Supported
Qwen2.5-7B-Instruct	qwen2.5-7b-instruct	Supported	Supported	Supported	Supported	Supported

Qwen-Plus-Character-2025-11-06	qwen-plus-character-2025-11-06	×	Supported	Supported	Supported	Supported

Visual understanding (Qwen-VL)

Model service	Model code	CPT full-parameter training (cpt)	SFT full-parameter training (sft)	SFT efficient training (efficient_sft)	DPO full-parameter training (dpo_full)	DPO efficient training (dpo_lora)
Qwen3-VL-8B-Instruct	qwen3-vl-8b-instruct	×	Supported	Supported	×	×
Qwen3-VL-8B-Thinking	qwen3-vl-8b-thinking	×	Supported	Supported	×	×
Qwen3-VL-4B-Instruct	qwen3-vl-4b-instruct	×	Supported	Supported	×	×

Qwen2.5-VL-72B-Instruct	qwen2.5-vl-72b-instruct	×	Supported	Supported	×	×
Qwen2.5-VL-32B-Instruct	qwen2.5-vl-32b-instruct	×	Supported	Supported	×	×
Qwen2.5-VL-7B-Instruct	qwen2.5-vl-7b-instruct	×	Supported	Supported	×	×

Comparison of tuning methods

Feature	CPT (Continual Pre-training)	SFT (Supervised Fine-tuning)	DPO (Direct Preference Optimization)
Summary	Supplements knowledge (Injects domain knowledge)	Learns to perform tasks (Follows instructions)	Performs tasks better (Aligns with human preferences)
Input data	10 million+ tokens Unlabeled domain text	Over 1,000 entries High-quality "question-answer" pairs	100+ sets "Better-worse" response pairs for the same instruction
Core objective	Domain adaptation. Learns specialized vocabulary and facts.	Teaches the model conversation formats and task execution capabilities.	Makes model outputs better align with human values and preferences.
Learning method	Self-supervised learning (Predicts the next word)	Supervised learning (Imitates the ground truth)	Direct preference learning (Increases the probability of good responses and decreases the probability of bad responses)
Model stage	Typically before SFT	After CPT and before DPO	Typically after SFT, as the final step for alignment.

Comparison of training patterns

Full-parameter training

Efficient training (LoRA, recommended)

Scenarios

• The model needs to acquire new capabilities

• Achieving optimal global performance.

• Optimizing model performance for specific scenarios.

• For cost-sensitive and time-sensitive scenarios.

Training time

Longer, with slower convergence.

Shorter, with faster convergence.

`hyper_parameters`: Supported settings

Supported parameters and their default values vary by model. To view specific default values, go to the console and select the model and training method.

Parameter	Recommended setting	Type	Description
`n_epochs` (Number of epochs) [Required]	Data size < 10,000: 3–5 Data size > 10,000: 1–2	Integer	The number of times the model iterates over the entire training set. Adjust this value based on your fine-tuning experience. More epochs increase training time and cost.
`learning_rate` (Learning rate)	Use the recommended default value provided by Model Studio.	Float	Controls the step-size for weight updates during training. A high learning rate can cause training to diverge, yielding poor results. A low learning rate can result in slow training and minimal performance improvement.
`freeze_vit` (Freeze visual backbone)	Adjust as needed	Boolean	Whether to freeze the visual backbone parameters, which prevents their weights from being updated during training. This parameter applies only to Qwen-VL (visual understanding) models. Warning Token-based billing is available only when freeze_vit is set to “true”.
`batch_size` (Batch size) [Required]	Use the recommended default value provided by Model Studio.	Integer	Specifies the number of training examples processed in one iteration. A small value can significantly increase training time. Default values vary by model; check the console for details.
`eval_steps` (Evaluation steps)	Adjust as needed	Integer	The step interval for evaluating the model's training accuracy and loss. This parameter affects how often `Validation Loss` and `Validation Token Accuracy` are displayed during fine-tuning.
`logging_steps` (Logging steps)	Adjust as needed	Integer	The step interval for logging fine-tuning progress.
`lr_scheduler_type` (Learning rate scheduler)	Recommended `linear`/`Inverse_sqrt`	String	The strategy for dynamically adjusting the learning rate during training. For details about each strategy, see Fine-tune a model in the console.
`max_length` (Sequence length) [Required]	8192	Integer	The maximum sequence length (in tokens) for a single training example. Examples that exceed this length are discarded. For the relationship between characters and tokens, see How to convert between tokens and characters.
`max_split_val_dataset_sample` (Max validation set samples)	Use the recommended default value provided by Model Studio.	Integer	When `"validation_file_ids"` is not set, the validation set that is automatically split by Model Studio contains a maximum of 1,000 entries. This parameter has no effect when `"validation_file_ids"` is set.
`split` (Training set ratio)	Use the recommended default value provided by Model Studio.	Float	If you do not set `"validation_file_ids"`, Alibaba Cloud Model Studio automatically uses 80% of the training file as the training set and 20% as the validation set. When `"validation_file_ids"` is set, this parameter has no effect.
`warmup_ratio` (Warm-up ratio)	Use the recommended default value provided by Model Studio.	Float	The proportion of the total training process used for learning rate warm-up. During warm-up, the learning rate linearly increases from a small initial value to the specified learning rate. This parameter helps stabilize training by limiting large parameter changes at the beginning. A ratio that is too high has an effect similar to a low learning rate, causing minimal performance changes. A ratio that is too low has an effect similar to a high learning rate and may degrade model performance. This parameter has no effect if the learning rate scheduler is set to `constant`.
`weight_decay` (Weight decay)	Use the recommended default value provided by Model Studio.	Float	The strength of L2 regularization. Regularization helps maintain the model's generalization ability. An excessively high value can reduce fine-tuning effectiveness.
Parameters for efficient fine-tuning (supports `efficient_sft` and `dpo_lora`) Note When you perform a second round of efficient fine-tuning on a model that has already been efficiently fine-tuned, the `lora_rank`, `lora_alpha`, and `lora_dropout` parameters must remain consistent.
`lora_rank` (LoRA rank)	64	Integer	The rank of the low-rank matrices in LoRA. A higher rank can improve fine-tuning results but may slightly increase training time.
`lora_alpha` (LoRA alpha)	Use the recommended default value provided by Model Studio.	Integer	The scaling factor that controls the combination of original model weights and the LoRA low-rank correction term. A larger alpha gives more weight to the LoRA correction, making the model rely more on task-specific information. A smaller alpha makes the model retain more knowledge from the base model.
`lora_dropout` (LoRA dropout)	Use the recommended default value provided by Model Studio.	Float	The dropout rate for the values in the low-rank matrices during LoRA training. Using the recommended value enhances the model's generalization capabilities. An overly large value can diminish the fine-tuning effect.
Parameters for publishing model parameter snapshots (for `efficient_sft` and `sft` only)
`save_strategy` (Snapshot save strategy)	It can be set to `epoch` or `steps`. When set to `steps`, you can adjust the save interval by setting the `save_steps` parameter.	String	The strategy for saving model parameter snapshots (checkpoints). Valid options are `epoch` (save after each epoch) and `steps` (save at a specified step interval).
`save_steps` (Save steps)	If you need to modify it manually, set it to an integer multiple of the `eval_steps` parameter.	Integer	The interval, in training steps, between each model parameter snapshot save.
`save_total_limit` (Snapshot save limit)	10	Integer	The maximum number of model parameter snapshots to retain. Once the limit is reached, older snapshots are automatically deleted.

Retrieve a fine-tuning job

To retrieve the details of a fine-tuning job, use the job_id returned when you create the job.

HTTP

curl 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<job_id>' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY} \
--header 'Content-Type: application/json'

Request parameters

Parameter	Type	Location	Required	Description
job_id	String	Path	Yes	The ID of the fine-tuning job.

Successful response

{
    "request_id": "d100cddb-ac85-4c82-bd5c-9b5421c5e94d",
    "output":
    {
        "job_id": "ft-202511272033-8ae7",
        "job_name": "ft-202511272033-8ae7",
        "status": "RUNNING",
        "finetuned_output": "qwen3-8b-ft-202511272033-8ae7",
        "model": "qwen3-8b",
        "base_model": "qwen3-8b",
        "training_file_ids":
        [
            "9e9ffdfa-c3bf-436e-9613-6f053c66aa6e"
        ],
        "validation_file_ids":
        [],
        "hyper_parameters":
        {
            "n_epochs": 3,
            "batch_size": 16,
            "max_length": 8192,
            "learning_rate": "1.6e-5",
            "lr_scheduler_type": "linear",
            "split": 0.9,
            "warmup_ratio": 0.05,
            "eval_steps": 50,
            "data_augmentation": true,
            "augmentation_ratio": "0.1,0.05,0.15",
            "augmentation_types": "dialogue_CN,general_purpose_CN,NLP",
            "save_strategy": "epoch",
            "save_total_limit": 10
        },
        "training_type": "sft",
        "create_time": "2025-11-27 20:33:15",
        "workspace_id": "llm-8v53etv3hwb8orx1",
        "user_identity": "1654290265984853",
        "modifier": "1654290265984853",
        "creator": "1654290265984853",
        "group": "llm",
        "max_output_cnt": 10
    }
}

Job status	Description
PENDING	The job is waiting to start.
QUEUING	The job is queued. Only one fine-tuning job runs at a time.
RUNNING	The job is running.
CANCELING	The job is being canceled.
SUCCEEDED	The job succeeded.
FAILED	The job failed.
CANCELED	The job was canceled.

Note

After a fine-tuning job succeeds, the finetuned_output field provides the resulting model ID. Use this ID for model deployment.

Get fine-tuning job logs

HTTP

curl 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<job_id>/logs?offset=0&line=1000' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY} \
--header 'Content-Type: application/json'

Use the offset and line parameters to retrieve a range of log lines. The offset parameter specifies the starting line, and the line parameter specifies the maximum number of lines to return.

Sample response:

{
    "request_id":"1100d073-4673-47df-aed8-c35b3108e968",
    "output":{
        "total":57,
        "logs":[
            "{Fine-tuning log 1}",
            "{Fine-tuning log 2}",
            ...
            ...
            ...
        ]
    }
}

Query and publish model checkpoints

Only SFT fine-tuning (efficient_sft and sft) supports saving and publishing checkpoints from intermediate training states.

List checkpoints for a fine-tuning job

curl 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<job_id>/checkpoints' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY} \
--header 'Content-Type: application/json'

Request parameters

Parameter	Type	Parameter location	Required	Description
job_id	String	Path Parameter	Yes	The ID of the fine-tuning job.

Sample response

Note

The checkpoint field contains the checkpoint ID, which specifies the checkpoint to publish in the Model publishing (optional) API. The model_name field contains the model ID used for model deployment. The finetuned_output field in the original fine-tuning job response is the model_name of the final checkpoint.

{
    "request_id": "c11939b5-efa6-4639-97ae-ed4597984647",
    "output":
    [
        {
            "create_time": "2025-11-11T16:25:42",
            "full_name": "ft-202511272033-8ae7-checkpoint-20",
            "job_id": "ft-202511272033-8ae7",
            "checkpoint": "checkpoint-20",
            "model_name": "qwen3-8b-instruct-ft-202511272033-8ae7",
            "status": "SUCCEEDED"
        }
    ]
}

Status	Description
PENDING	The checkpoint is pending publication. You must publish it using the Model publishing API before you can use it for model deployment and invocation.
PROCESSING	The checkpoint is being published.
SUCCEEDED	The checkpoint has been published successfully. You can now use it for model deployment and invocation.
FAILED	The checkpoint failed to publish.

Model publishing (optional)

Note

In Model Studio, after a fine-tuning job completes, you must export a checkpoint before you can deploy the model.

Exported checkpoints are stored in cloud storage. You cannot access or download them at this time.

curl --request GET 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<job_id>/export/<checkpoint_id>?model_name=<model_name>' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY} \
--header 'Content-Type: application/json'

Request parameters

Parameter	Type	Parameter location	Required	Description
job_id	String	Path Parameter	Yes	The ID of the fine-tuning job.
checkpoint_id	String	Path Parameter	Yes	The ID of the checkpoint to publish.
model_name	String	Path Parameter	Yes	The custom model ID to assign to the published model.

Sample response

{
    "request_id": "ed3faa41-6be3-4271-9b83-941b23680537",
    "output": true
}

The publishing task is asynchronous. Use the List checkpoints for a fine-tuning job API to monitor the publishing status of the checkpoint.

More fine-tuning operations

List fine-tuning jobs

HTTP

curl 'https://dashscope.aliyuncs.com/api/v1/fine-tunes' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY} \
--header 'Content-Type: application/json'

Cancel a fine-tuning job

Cancels a running fine-tuning job.

HTTP

curl --request POST 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<job_id>/cancel' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY} \
--header 'Content-Type: application/json'

Delete a fine-tuning job

You cannot delete a running fine-tuning job.

HTTP

curl --request DELETE 'https://dashscope.aliyuncs.com/api/v1/fine-tunes/<job_id>' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY} \
--header 'Content-Type: application/json'

Model deployment and invocation

Model deployment

To deploy the model, go to the model deployment console.

Model invocation

Once the model deployment status is RUNNING, you can invoke the fine-tuned model just like any other model.

You can also get the Model Code from the model deployment console.

For details on usage and parameters, see the DashScope API Reference.

HTTP

curl 'https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation' \
--header 'Authorization: Bearer '${DASHSCOPE_API_KEY}  \
--header 'Content-Type: application/json' \
--data '{
    "model": "<your_model_instance_id>",
    "input":{
        "messages":[
            {
                "role": "user",
                "content": "Who are you?"
            }
        ]
    },
    "parameters": {
        "result_format": "message"
    }
}'

Prerequisites

Tuning file upload

Preparing fine-tuning files

SFT training set

Upload fine-tuning file to Model Studio

OpenAI-compatible Files API

Model fine-tuning

Create a fine-tuning job

HTTP

Parameters

Response

Base models (model) and training types (training_type)

Supported models

Singapore

North China 2 (Beijing)

Comparison of tuning methods

Comparison of training patterns

hyper_parameters: Supported settings

Retrieve a fine-tuning job

HTTP

Request parameters

Successful response

Get fine-tuning job logs

HTTP

Query and publish model checkpoints

List checkpoints for a fine-tuning job

Model publishing (optional)

More fine-tuning operations

List fine-tuning jobs

Cancel a fine-tuning job

Delete a fine-tuning job

Model deployment and invocation

Model deployment

Model invocation

Base models (`model`) and training types (`training_type`)

`hyper_parameters`: Supported settings