My Models

In My Models, you can manage models fine-tuned in Alibaba Cloud Model Studio and models imported from OSS (LoRA fine-tuning and full-parameter fine-tuning only), covering first-time OSS authorization, model file preparation and validation, import form completion, model status management, and common issue troubleshooting. Supports cloud deployment with high-concurrency, low-latency inference services.

Supported Models

On the My Models page, you can manage two types of models:

• Fine-tuned models: Models fine-tuned in Alibaba Cloud Model Studio. For more information, see Supported Models for Model Fine-tuning.

• Imported models: Models imported from Object Storage Service (OSS). This supports importing models that underwent LoRA fine-tuning or full-parameter fine-tuning.

  Importing full-parameter fine-tuned models is a whitelist feature. To enable it, contact your account manager.

  Model list:

  China (Beijing)

  Supported Models	Model Name
  Qwen3	Qwen3-32B
  Qwen3	Qwen3-14B
  Qwen3	Qwen3-8B
  Qwen3	Qwen3-4B-Instruct-2507
  Qwen3-VL	Qwen3-VL-8B-Instruct
  Qwen2.5	Qwen2.5-72B-Instruct
  Qwen2.5	Qwen2.5-32B-Instruct
  Qwen2.5	Qwen2.5-14B-Instruct
  Qwen2.5	Qwen2.5-7B-Instruct
  Qwen2.5-VL	Qwen2.5-VL-72B-Instruct
  Qwen2.5-VL	Qwen2.5-VL-7B-Instruct

  Singapore

  Supported Models	Model Name
  -	-

  Data updates may be delayed. Refer to the actual display on the interface when you import models.

Import Models

This section describes how to import models from OSS using the Alibaba Cloud Model Studio console.

Importing models is not supported through API or command line operations.

1. On the My Models page, click Import Model to open the Import Model interface.

2. Enter a Model Name and select the Base Model (the model to be fine-tuned).

3. For Import Method, select Import from OSS. Currently, no other methods are supported. You need to manually upload the associated files for your LoRA fine-tuned or full-parameter fine-tuned model to Alibaba Cloud Object Storage Service (OSS). Upload View model file examples

   Important

   • For the first import, complete the authorization as prompted on the interface and add a tag to the target bucket. For more information, see How to import files from OSS to Alibaba Cloud Model Studio for the first time.
   • Supported OSS bucket storage classes do not include Archive, Cold Archive, or Deep Cold Archive. Encrypted buckets are supported. Private buckets are supported.
   • Accessing files in the root directory of an OSS bucket is not supported. Select an existing subdirectory or create a new one within the OSS bucket for Alibaba Cloud Model Studio to access.
   • You can import model files of any size. After import, they use the free storage space provided by Alibaba Cloud Model Studio.

4. After clicking OK, the system begins importing the model. During peak request times, this process may take longer. You can wait patiently.

Prepare LoRA Model Files

Before importing, place the LoRA model files in a subdirectory of an OSS Bucket per the following requirements (the Bucket root directory is not supported), and pass the system's automatic validation before submission. The model files must be placed directly under the selected subdirectory; the system detects them automatically.

Required files and directory structure

The subdirectory must contain the following files: adapter_model.safetensors (LoRA adapter weights, SafeTensors format), adapter_config.json (configuration file containing rank, alpha, and other parameters), and config.json (base model configuration). After selecting a directory, the system automatically validates the format and completeness of these files.

Training parameter constraints

• rank value: rank must be 8, 16, 32, or 64, and all LoRA layers of the same model must use the same rank value; otherwise the model cannot be imported.
• Vocabulary cannot be modified: Models that added new tokens or modified the original vocabulary during training cannot be imported; the vocabulary must exactly match the base model's.
• Chat template cannot be modified: Models that modified the chat_template during training cannot be imported; it must match the base model's default configuration. The chat_template is located in the chat_template field of config.json or tokenizer_config.json.
• Vision models must freeze VIT: Vision-language models must freeze the Vision Transformer part. If the LoRA adapter contains visual-related weight parameters (i.e., VIT not frozen), the model cannot be imported.

You can run the following script before importing to check whether adapter_model.safetensors contains parameter keys starting with visual, to determine whether VIT is frozen.

from safetensors import safe_open
import argparse

def print_safetensor_structure(file_path):
    print(f"Loading safetensor file: {file_path}")
    print("="*80)

    with safe_open(file_path, framework="pt") as f:
        keys = f.keys()
        print(f"Found {len(keys)} tensors in the file:\n")

        for key in sorted(keys):
            tensor = f.get_tensor(key)
            shape = tuple(tensor.shape)
            dtype = str(tensor.dtype)
            device = tensor.device if hasattr(tensor, 'device') else 'cpu'

            lora_tag = " [LoRA]" if "lora_A" in key or "lora_B" in key else ""

            print(f"[{dtype:>14}] {shape} | {key} {lora_tag}")

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Print structure of a .safetensors LoRA adapter.")
    parser.add_argument("filepath", type=str, help="Path to the .safetensors file")
    args = parser.parse_args()

    print_safetensor_structure(args.filepath)

Method: If the script output contains parameter keys starting with visual (such as visual.encoder.layer.0...), the VIT part is not frozen and the model cannot be imported; if it only contains LoRA-related keys such as lora_A and lora_B, the VIT is frozen and the model can be imported.

Pre-submission automatic validation

On the import page, after selecting a model directory, the system automatically calls the file validation interface to check the format and completeness of the model files in the directory. If validation fails, a red prompt appears below the directory field and blocks submission. Fix the files per the prompt before submitting. A common failure cause is missing required files, with error code AvailableModelFileNotFound.

Import Form Fields

The import model form fields are as follows:

Field	Description	Constraint
Model name	Enter a display name for the model.	Required, up to 50 characters
Base model	Select the base model used for LoRA training; must match the training base.	Required, dropdown select
Training method	Options depend on the selected base model; after selecting a base model, options render automatically with the first one selected by default.	Required, dropdown select
Import source	Only "Import from OSS" is supported; no other options.	Read-only, selected by default
Bucket	Select the OSS bucket storing the model files; only buckets with the bailian-datahub-access=read tag are listed.	Required, dropdown select
Model directory	Browse and select the subdirectory containing the model checkpoint in the chosen bucket; the bucket root directory is not selectable.	Required, tree select
Model encryption	The platform automatically enables OSS server-side encryption (SSE-OSS) for exported model files, using fully managed OSS keys with AES256 as the encryption algorithm.	Read-only, platform-enforced

Imported model states include Creating (importing), Created (ready to deploy), Creation Failed (import failed), and Expired (source files changed).

What to do next

Manage My Models

On the My Models page, you can view all fine-tuned and imported models in the current workspace. You can also perform operations such as deployment and deletion.

Delete: Undeploy models before deleting them. If the model comes from Model Fine-tuning, this operation does not delete records from the Model Fine-tuning interface.

Status and transitions

Imported model states include Creating, Created, Creation Failed, and Expired. Creating means importing is in progress; Created means the import is complete and deployable; Creation Failed means the import was unsuccessful; Expired means the OSS source model files changed after Created. The list silently auto-refreshes every 3 seconds for models in the Creating state, which is normal behavior, not an API exception.

The Creation Failed status has a "Details" link next to it; hover to view the failure error code (such as AvailableModelFileNotFound) and the corresponding oss://bucket/path, to help locate the failed file.

The Expired status can be hovered to view a pop-up showing "Changes detected in the following files" and the list of changed source file names, prompting you to re-import.

Actions column availability

Each row's actions column provides Deploy, Incremental training, and Delete buttons. Button availability depends on the model status:

• Deploy: Only available for the Created state. Clicking it goes to the deployment creation page; for deployment operations, see Deployment and O&M. The button is unavailable for other states or when no deployment method is available.
• Incremental training: Models imported from OSS do not support incremental training; the button is unavailable.
• Delete: Models in the Creating state cannot be deleted; quantized models must be deleted on the model compression page.

After clicking Delete in the actions column and confirming, only the Model Studio-side model record is removed; OSS source files are not affected. The search box at the top of the list filters models by name and supports clearing/resetting. The Source column renders by import source: OSS imports show oss://bucket/path, training jobs show the source job ID (if deleted, shows "Training job deleted"). The Supported deployment methods column displays the model's supported deployment methods and training method tags (such as full-parameter, LoRA, or quantization tags) and billing methods (such as per-token billing, per-model-unit billing); shows "-" when no deployment method is available.

All operations are performed on a single model; batch deletion or batch deployment is not supported.

Invoke My Models

Models provide inference services only after successful deployment (API invocation only). For specific operations and billing methods, see Invoke After Deployment.

• Key restrictions: After you successfully deploy models fine-tuned in Alibaba Cloud Model Studio, invoke them only using the API key of their workspace. Currently, only DashScope invocation is supported, not (OpenAI compatible) invocation.
• Optimize inference results: When invoking, explicitly configure inference hyperparameters (such as presence_penalty, frequency_penalty, and repetition_penalty). Ensure these settings match the inference hyperparameters used during your offline fine-tuning.

Model File Examples

• LoRA fine-tuning typically generates two checkpoint files: best_model_checkpoint (the checkpoint with the best performance on the validation set) and last_model_checkpoint (the last saved checkpoint). You can import either one as needed.

  Ensure the imported model files include adapter_model.safetensors and adapter_config.json. To learn about LoRA fine-tuning methods, refer to the model fine-tuning section in the Alibaba Cloud Large Language Model ACP Course.

  Examples of model files (checkpoint files) to prepare:

  

• Files to import for full-parameter fine-tuning must be consistent with open source models.

FAQ

How do I import files from OSS to Alibaba Cloud Model Studio for the first time?

If you import files from OSS to Alibaba Cloud Model Studio for the first time, complete the authorization as prompted on the interface and add the bailian-datahub-access tag to the target OSS bucket before you import.

If you are unfamiliar with the concepts and differences between root accounts and RAM users, first read Permission Management.

Use a Root Account

1. Click Go to Authorization.

   

2. In the dialog box, click Confirm Authorization. The system automatically enables the OSS Service-linked Role for you (a prerequisite).

   This usually takes effect within seconds. During peak service hours, there may be a slight delay.

   What to do if "This request failed. Try resubmitting or contact the administrator. Error code: 10041495" appears.

   

3. Add the bailian-datahub-access tag to the target OSS bucket.

   This tag marks buckets accessible by Alibaba Cloud Model Studio. Alibaba Cloud Model Studio cannot access untagged buckets.

   1. Go to the OSS Management Console. In the navigation pane on the left, click Bucket List. The buckets that you created are displayed.

   2. In the Tag column for the target bucket, hover over the icon, then click Go to Edit.

      

   3. Click Create Tag.

      

   4. Click Tag. Set the tag name to bailian-datahub-access and the tag value to read. Then, click Save.

      

4. Return to the Import Model interface, reselect the target bucket, and try importing again.

   Note that Alibaba Cloud Model Studio does not support accessing files stored in the root directory of a bucket. Select an existing folder or create a new one within the bucket for Alibaba Cloud Model Studio to access.

Use a RAM User

1. Click Go to Authorization.

   

2. In the dialog box that appears, click Confirm Authorization. The messages Authorization Failed and The Current User Does Not Have Permission To Create A Service-linked Role are displayed. This error occurs because the current RAM user does not have permission to create a service-linked role. To proceed, you must first grant the RAM user permission to create a service-linked role, and then grant the RAM user permission to access OSS through Alibaba Cloud Model Studio.

   

3. Grant the RAM user permission to create a service-linked role.

   1. Log on to the RAM console as the root account. In the left navigation pane, choose Permission Management > Permission Policies. Then click Create Policy on the interface.
   2. In the Script Editor, you can enter the content from the following script for Effect, Action, Resource, and Condition. Then click OK.

{
    "Action": [
        "ram:CreateServiceLinkedRole"
    ],
    "Resource": "*",
    "Effect": "Allow",
    "Condition": {
        "StringEquals": {
            "ram:ServiceName": "datahub.sfm.aliyuncs.com"
        }
    }
}

3. After you enter the policy name, click OK.

   

4. In the navigation pane on the left, choose Identity Management > Users. Find the RAM user that you want to authorize. In the Actions column for that user, click Add Permissions.

5. In the permission policy, select the policy you just created (custom policy). Click Confirm New Authorization. The RAM user now has permissions to create a service-linked role.

   

6. Authorize the RAM user to access OSS through Alibaba Cloud Model Studio.

   1. Return to the Import Model interface and click Go to Authorization.

      

   2. In the dialog box, click Confirm Authorization. This action automatically enables the OSS Service-linked Role, which is a prerequisite.

      This usually takes effect within seconds. During peak service hours, there may be a slight delay.

      What to do if "This request failed. Try resubmitting or contact the administrator. Error code: 10041495" appears.

      

7. Add the bailian-datahub-access tag to the target OSS bucket.

   This tag marks buckets accessible by Alibaba Cloud Model Studio. Alibaba Cloud Model Studio cannot access untagged buckets.

   1. Go to the OSS Management Console. In the navigation pane on the left, click Bucket List. View your created buckets.

   2. In the Tag column of the bucket you want to tag, hover over the icon, then click Go to Edit.

      

   3. You can click Create Tag.

      

   4. Click Tag. Add a tag with the tag name bailian-datahub-access and tag value read. Then click Save.

      

8. Return to the Import Model interface. Reselect the target bucket and try importing again.

   Note that Alibaba Cloud Model Studio does not support accessing files stored in the root directory of a bucket. Select an existing folder or create a new one within the bucket for Alibaba Cloud Model Studio to access.

What to do if error "10041495" occurs?

This usually occurs because the root account has not enabled Object Storage Service (OSS). Follow these steps:

1. The root account must go to the OSS Management Console. Enable OSS as prompted on the interface.
2. Return to the Alibaba Cloud Model Studio Import Model interface. Then, try authorizing again.

What to do when import fails with AvailableModelFileNotFound?

This error means the model directory file validation did not pass (a format or integrity issue) and cannot be resolved by simply re-uploading the files. Check that the selected directory contains valid adapter_model.safetensors, adapter_config.json, and config.json, confirm that the rank, vocabulary, and chat_template constraints are all met, then reselect the directory and submit.

What does the Expired model status mean?

Expired means that after the model was Created, the OSS source model files changed. This is a normal detection behavior, not a fault. Hover over the Expired status to view the list of changed file names; you must re-import the model to restore it to a usable state.

What to do when a target Bucket is not selectable in the Bucket dropdown?

This is an authorization requirement, not a fault. Under the new authorization method, Buckets without the bailian-datahub-access=read tag are not selectable in the dropdown. Go to the OSS management console to add the tag to the target Bucket, then return to the import page and reselect.

Is the My Models list auto-refreshing every few seconds a fault?

No. When the list detects models in the Creating state, it silently refreshes every 3 seconds to fetch the latest status, and stops automatically when no Creating state exists. This is normal behavior.

Why does the imported model produce different inference results compared to local vLLM or SGLang?

For inference parameter alignment, see API deployment guide.

Does deleting an imported model affect the source files in OSS?

No. Delete only removes the Model Studio-side model record; you must re-import to recover. OSS source files belong to you; Model Studio only reads them via the bailian-datahub-access=read tag, and deleting the model does not modify any files in OSS.

How to upgrade from the legacy authorization method?

If you previously used the legacy authorization method, the Bucket field on the import page shows "We recommend converting to the new Bucket authorization method to improve security" with a "Convert directly" link. Click it to open a confirmation dialog; confirm to upgrade to the service-linked role authorization method. The upgrade does not affect existing data.