Skip to content

update #57

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tastelikefeet merged 1 commit into from
Feb 13, 2026
+49 −53
Merged

update #57

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
73 changes: 38 additions & 35 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -51,7 +51,7 @@ be reused in [ms-swift](https://github.com/modelscope/ms-swift).
pip install 'twinkle-kit'
```

### Installation from Source:
### Install from Source:

```shell
git clone https://github.com/modelscope/twinkle.git
Expand All @@ -75,13 +75,14 @@ pip install -e .

## Changelog

- 🎉2026-02-10 Initial version of Twinkle✨ released, including SFT/PT/RL for text models and serverless training capabilities on [ModelScope](https://modelscope.cn).
- 🎉2026-02-13 Initial version of Twinkle✨ released, including SFT/PT/RL support for text models and serverless training capabilities on [ModelScope](https://modelscope.cn).

# ModelScope Community
## Training as a Service on ModelScope

## ModelScope Official Environment

The ModelScope community provides an official environment for running Twinkle. The API endpoint is: [base_url](https://www.modelscope.cn/twinkle). Developers can refer to our [documentation](docs/source_en/Usage%20Guide/ModelScope-Official-Resources.md) for usage instructions.
We are rolling out training service built atop Twinkle✨ on ModelScope. It is currently in _Beta_. You may
sign up for free access by joining the [Twinkle-Explorers](https://modelscope.cn/organization/twinkle-explorers) organization, and
train via API endpoint `base_url=https://www.modelscope.cn/twinkle`. For more details, please refer to
our [documentation](docs/source_en/Usage%20Guide/ModelScope-Official-Resources.md).
Comment on lines +82 to +85
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist bot Feb 13, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The linked English documentation file (docs/source_en/Usage Guide/ModelScope-Official-Resources.md) appears to be outdated. It still contains the old instructions, which are inconsistent with the new process described in this README and the updated Chinese documentation (docs/source_zh/使用指引/训练服务.md). To avoid confusion for English-speaking users, please consider updating this document and renaming it (e.g., to Training-Service.md) to match the changes made to the Chinese documentation.


## Supported Hardware

Expand All @@ -95,29 +96,33 @@ The ModelScope community provides an official environment for running Twinkle. T
## Supported Models

We will be adding support for more models as new models are released. The following table lists current models
supported on Twinkle✨ framework. However, the models supported on our serverless training backend may be a
much smaller subset. Please refer to the [doc](link) section for more information.

| Model Type | Model ID on[ModelScope](https://modelscope.cn) | Requires | Megatron Support | HF Model ID |
| ------------------- | --------------------------------------------------------------------------------------------------------------------- | -------------------- | ---------------- | ---------------------------------------------------------------------------------------------------------- |
| qwen3 series | [Qwen/Qwen3-0.6B-Base](https://modelscope.cn/models/Qwen/Qwen3-0.6B-Base)~32B | transformers>=4.51 | ✔ | [Qwen/Qwen3-0.6B-Base](https://huggingface.co/Qwen/Qwen3-0.6B-Base) |
| qwen3_moe series | [Qwen/Qwen3-30B-A3B-Base](https://modelscope.cn/models/Qwen/Qwen3-30B-A3B-Base) | transformers>=4.51 | ✔ | [Qwen/Qwen3-30B-A3B-Base](https://huggingface.co/Qwen/Qwen3-30B-A3B-Base) |
| | [Qwen/Qwen3-30B-A3B](https://modelscope.cn/models/Qwen/Qwen3-30B-A3B)~235B | transformers>=4.51 | ✔ | [Qwen/Qwen3-30B-A3B](https://huggingface.co/Qwen/Qwen3-30B-A3B) |
| qwen2 series | [Qwen/Qwen2-0.5B-Instruct](https://modelscope.cn/models/Qwen/Qwen2-0.5B-Instruct) ~72B | transformers>=4.37 | ✔ | [Qwen/Qwen2-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) |
| | [Qwen/Qwen2.5-0.5B-Instruct](https://modelscope.cn/models/Qwen/Qwen2.5-0.5B-Instruct)~72B | transformers>=4.37 | ✔ | [Qwen/Qwen2.5-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct) |
| | [Qwen/Qwen2.5-0.5B](https://modelscope.cn/models/Qwen/Qwen2.5-0.5B)~72B | transformers>=4.37 | ✔ | [Qwen/Qwen2.5-0.5B](https://huggingface.co/Qwen/Qwen2.5-0.5B) |
| qwen2_moe series | [Qwen/Qwen1.5-MoE-A2.7B-Chat](https://modelscope.cn/models/Qwen/Qwen1.5-MoE-A2.7B-Chat) | transformers>=4.40 | ✔ | [Qwen/Qwen1.5-MoE-A2.7B-Chat](https://huggingface.co/Qwen/Qwen1.5-MoE-A2.7B-Chat) |
supported on Twinkle✨ framework.

>[!Note]
> For serverless training service accessed via `base_url=https://www.modelscope.cn/twinkle`, it currently supports
> one training base at a time, and currently it is [Qwen3-30B-A3B-Instruct-2507](https://modelscope.cn/models/Qwen/Qwen3-30B-A3B-Instruct-2507).


| Model Type | Model ID on [ModelScope](https://modelscope.cn) | Requires | Megatron Support | HF Model ID |
| ------------------- |--------------------------------------------------------------------------------------------------------------------------| -------------------- | ---------------- | ---------------------------------------------------------------------------------------------------------- |
Comment on lines +106 to +107
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist bot Feb 13, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The markdown table formatting appears to be broken. The separator for the second column is not aligned and lacks spacing, which can cause rendering issues on some platforms. Please adjust the separator line to align with the headers for better readability and consistent rendering.

Suggested change
| Model Type | Model ID on [ModelScope](https://modelscope.cn) | Requires | Megatron Support | HF Model ID |
| ------------------- |--------------------------------------------------------------------------------------------------------------------------| -------------------- | ---------------- | ---------------------------------------------------------------------------------------------------------- |
| Model Type | Model ID on [ModelScope](https://modelscope.cn) | Requires | Megatron Support | HF Model ID |
| ------------------- | ------------------------------------------------------------------------------------------------------------------------ | -------------------- | ---------------- | ---------------------------------------------------------------------------------------------------------- |

| qwen3 series | [Qwen/Qwen3-0.6B-Base](https://modelscope.cn/models/Qwen/Qwen3-0.6B-Base)~32B | transformers>=4.51 | ✅ | [Qwen/Qwen3-0.6B-Base](https://huggingface.co/Qwen/Qwen3-0.6B-Base) |
| qwen3_moe series | [Qwen/Qwen3-30B-A3B-Base](https://modelscope.cn/models/Qwen/Qwen3-30B-A3B-Base) | transformers>=4.51 | ✅ | [Qwen/Qwen3-30B-A3B-Base](https://huggingface.co/Qwen/Qwen3-30B-A3B-Base) |
| | [Qwen/Qwen3-30B-A3B](https://modelscope.cn/models/Qwen/Qwen3-30B-A3B)~235B | transformers>=4.51 | ✅ | [Qwen/Qwen3-30B-A3B](https://huggingface.co/Qwen/Qwen3-30B-A3B) |
| qwen2 series | [Qwen/Qwen2-0.5B-Instruct](https://modelscope.cn/models/Qwen/Qwen2-0.5B-Instruct) ~72B | transformers>=4.37 | ✅ | [Qwen/Qwen2-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2-0.5B-Instruct) |
| | [Qwen/Qwen2.5-0.5B-Instruct](https://modelscope.cn/models/Qwen/Qwen2.5-0.5B-Instruct)~72B | transformers>=4.37 | ✅ | [Qwen/Qwen2.5-0.5B-Instruct](https://huggingface.co/Qwen/Qwen2.5-0.5B-Instruct) |
| | [Qwen/Qwen2.5-0.5B](https://modelscope.cn/models/Qwen/Qwen2.5-0.5B)~72B | transformers>=4.37 | ✅ | [Qwen/Qwen2.5-0.5B](https://huggingface.co/Qwen/Qwen2.5-0.5B) |
| qwen2_moe series | [Qwen/Qwen1.5-MoE-A2.7B-Chat](https://modelscope.cn/models/Qwen/Qwen1.5-MoE-A2.7B-Chat) | transformers>=4.40 | ✅ | [Qwen/Qwen1.5-MoE-A2.7B-Chat](https://huggingface.co/Qwen/Qwen1.5-MoE-A2.7B-Chat) |
| chatglm4 series | [ZhipuAI/glm-4-9b-chat](https://modelscope.cn/models/ZhipuAI/glm-4-9b-chat) | transformers>=4.42 | ✘ | [zai-org/glm-4-9b-chat](https://huggingface.co/zai-org/glm-4-9b-chat) |
| | [ZhipuAI/LongWriter-glm4-9b](https://modelscope.cn/models/ZhipuAI/LongWriter-glm4-9b) | transformers>=4.42 | ✘ | [zai-org/LongWriter-glm4-9b](https://huggingface.co/zai-org/LongWriter-glm4-9b) |
| glm_edge series | [ZhipuAI/glm-edge-1.5b-chat](https://modelscope.cn/models/ZhipuAI/glm-edge-1.5b-chat) | transformers>=4.46 | ✘ | [zai-org/glm-edge-1.5b-chat](https://huggingface.co/zai-org/glm-edge-1.5b-chat) |
| | [ZhipuAI/glm-edge-4b-chat](https://modelscope.cn/models/ZhipuAI/glm-edge-4b-chat) | transformers>=4.46 | ✘ | [zai-org/glm-edge-4b-chat](https://huggingface.co/zai-org/glm-edge-4b-chat) |
| internlm2 series | [Shanghai_AI_Laboratory/internlm2-1_8b](https://modelscope.cn/models/Shanghai_AI_Laboratory/internlm2-1_8b) | transformers>=4.38 | ✘ | [internlm/internlm2-1_8b](https://huggingface.co/internlm/internlm2-1_8b) |
| | [Shanghai_AI_Laboratory/internlm2-chat-7b](https://modelscope.cn/models/Shanghai_AI_Laboratory/internlm2-chat-7b) | transformers>=4.38 | ✘ | [internlm/internlm2-chat-7b](https://huggingface.co/internlm/internlm2-chat-7b) |
| deepseek_v1 | [deepseek-ai/deepseek-vl-7b-chat](https://modelscope.cn/models/deepseek-ai/deepseek-vl-7b-chat) | transformers>=4.39.4 | ✔ | —— |
| | [deepseek-ai/DeepSeek-V2-Lite](https://modelscope.cn/models/deepseek-ai/DeepSeek-V2-Lite) | transformers>=4.39.3 | ✔ | [deepseek-ai/DeepSeek-V2-Lite](https://huggingface.co/deepseek-ai/DeepSeek-V2-Lite) |
| | [deepseek-ai/DeepSeek-V2.5](https://modelscope.cn/models/deepseek-ai/DeepSeek-V2.5) | transformers>=4.39.3 | ✔ | [deepseek-ai/DeepSeek-V2.5](https://huggingface.co/deepseek-ai/DeepSeek-V2.5) |
| | [deepseek-ai/DeepSeek-R1](https://modelscope.cn/models/deepseek-ai/DeepSeek-R1) | transformers>=4.39.3 | ✔ | [deepseek-ai/DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1) |
| deepSeek-r1-distill | [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://modelscope.cn/models/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B) ~32B | transformers>=4.37 | ✔ | [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B) |
| deepseek_v1 | [deepseek-ai/deepseek-vl-7b-chat](https://modelscope.cn/models/deepseek-ai/deepseek-vl-7b-chat) | transformers>=4.39.4 | ✅ | —— |
| | [deepseek-ai/DeepSeek-V2-Lite](https://modelscope.cn/models/deepseek-ai/DeepSeek-V2-Lite) | transformers>=4.39.3 | ✅ | [deepseek-ai/DeepSeek-V2-Lite](https://huggingface.co/deepseek-ai/DeepSeek-V2-Lite) |
| | [deepseek-ai/DeepSeek-V2.5](https://modelscope.cn/models/deepseek-ai/DeepSeek-V2.5) | transformers>=4.39.3 | ✅ | [deepseek-ai/DeepSeek-V2.5](https://huggingface.co/deepseek-ai/DeepSeek-V2.5) |
| | [deepseek-ai/DeepSeek-R1](https://modelscope.cn/models/deepseek-ai/DeepSeek-R1) | transformers>=4.39.3 | ✅ | [deepseek-ai/DeepSeek-R1](https://huggingface.co/deepseek-ai/DeepSeek-R1) |
| deepSeek-r1-distill | [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://modelscope.cn/models/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B) ~32B | transformers>=4.37 | ✅ | [deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B](https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B) |

For a more detailed model support list 👉 [Quick Start.md](https://github.com/modelscope/twinkle/blob/dev/docs/source/%E4%BD%BF%E7%94%A8%E6%8C%87%E5%BC%95/%E5%BF%AB%E9%80%9F%E5%BC%80%E5%A7%8B.md)

Expand All @@ -141,18 +146,20 @@ twinkle.initialize(mode='ray', groups=device_group, global_device_mesh=device_me


def train():
# to load model from Hugging Face, use 'hf://...'
base_model = 'ms://Qwen/Qwen2.5-7B-Instruct'
# 1000 samples
dataset = Dataset(dataset_meta=DatasetMeta('ms://swift/self-cognition', data_slice=range(1000)))
# Set template to prepare encoding
dataset.set_template('Template', model_id='ms://Qwen/Qwen2.5-7B-Instruct')
dataset.set_template('Template', model_id=base_model)
# Preprocess the dataset to standard format
dataset.map(SelfCognitionProcessor('twinkle LLM', 'ModelScope Community'))
# Encode dataset
dataset.encode()
# Global batch size = 8, for GPUs, so 1 sample per GPU
dataloader = DataLoader(dataset=dataset, batch_size=8, min_batch_size=8)
# Use a TransformersModel
model = TransformersModel(model_id='ms://Qwen/Qwen2.5-7B-Instruct', remote_group='default')
model = TransformersModel(model_id=base_model, remote_group='default')

lora_config = LoraConfig(
r=8,
Expand Down Expand Up @@ -184,7 +191,7 @@ if __name__ == '__main__':
train()
```

### Tinker-Like Remote API
### Using Tinker-Like API

```python
import os
Expand All @@ -196,17 +203,19 @@ from twinkle.dataset import Dataset, DatasetMeta
from twinkle.preprocessor import SelfCognitionProcessor
from twinkle.server.tinker.common import input_feature_to_datum

base_model = "Qwen/Qwen2.5-0.5B-Instruct"
base_model = 'ms://Qwen/Qwen2.5-0.5B-Instruct'
base_url='http://www.modelscope.cn/twinkle'
api_key=os.environ.get('MODELSCOPE_TOKEN')

# Use twinkle dataset to load the data
dataset = Dataset(dataset_meta=DatasetMeta('ms://swift/self-cognition', data_slice=range(500)))
dataset.set_template('Template', model_id=f'ms://{base_model}', max_length=256)
dataset.set_template('Template', model_id=base_model, max_length=256)
dataset.map(SelfCognitionProcessor('twinkle Model', 'twinkle Team'), load_from_cache_file=False)
dataset.encode(batched=True, load_from_cache_file=False)
dataloader = DataLoader(dataset=dataset, batch_size=8)

# Initialize tinker client
service_client = init_tinker_compat_client(base_url='http://www.modelscope.cn/twinkle', api_key=os.environ.get('MODELSCOPE_SDK_TOKEN'))
service_client = init_tinker_compat_client(base_model, api_key)
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist bot Feb 13, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

There seems to be a bug in this example. The init_tinker_compat_client function is called with base_model as the first argument, but it likely expects base_url. The base_url variable is defined on line 207 but never used. This will probably cause the example code to fail. Please verify the function signature and pass base_url instead.

Suggested change
service_client = init_tinker_compat_client(base_model, api_key)
service_client = init_tinker_compat_client(base_url, api_key)

training_client = service_client.create_lora_training_client(base_model=base_model, rank=16)

# Training loop: use input_feature_to_datum to transfer the input format
Expand All @@ -223,12 +232,6 @@ for epoch in range(3):
training_client.save_state(f"twinkle-lora-{epoch}").result()
```

Launch training:

```shell
python3 train.py
```

## Architecture Design

<img src="assets/framework.jpg" style="max-width: 500px; width: 100%;" />
Expand Down
2 changes: 1 addition & 1 deletion README_ZH.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,7 +69,7 @@ pip install -e .

## 魔搭社区官方环境

魔搭社区提供了Twinkle运行的官方环境，调用端点为:[base_url](https://www.modelscope.cn/twinkle)，开发者可以参考我们的[文档](docs/source_zh/使用指引/魔搭官方环境.md)来进行使用。
魔搭社区提供了Twinkle运行的官方环境，调用端点为:[base_url](https://www.modelscope.cn/twinkle)，开发者可以参考我们的[文档](docs/source_zh/使用指引/训练服务.md)来进行使用。

## 支持的硬件

Expand Down
6 changes: 0 additions & 6 deletions docs/source_zh/使用指引/快速开始.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,17 +1,12 @@
<div align="center">

## ✨ Twinkle 是什么？

大模型训练组件库。基于 PyTorch，更简洁、更灵活、生产就绪。

<p align="center">
🧩 <b>松耦合架构</b> · 标准化接口<br>
🚀 <b>多运行模式</b> · torchrun / Ray / HTTP<br>
🔌 <b>多框架兼容</b> · Transformers / Megatron<br>
👥 <b>多租户支持</b> · 单基座模型部署
</p>

</div>

## Twinkle 适配性

Expand All @@ -23,7 +18,6 @@ Twinkle 和 [ms-swift](https://github.com/modelscope/ms-swift) 都是模型训
- 如果你是大模型研究者，希望定制模型或训练方法
- 如果你善于编写 training loop，希望定制训练过程
- 如果你希望提供企业级或商业化训练平台
- 如果你缺少训练硬件，希望使用社区资源

### 何时选择ms-swift

Expand Down
21 changes: 10 additions & 11 deletions docs/source_zh/使用指引/魔搭官方环境.md → docs/source_zh/使用指引/训练服务.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,24 +1,23 @@
# 魔搭官方环境
# ModelScope上的Twinkle训练服务

在 Twinkle 框架开源的同时，我们在[魔搭社区官网](https://www.modelscope.cn)上提供了样例训练资源。开发者可以通过ModelScope Token进行训练。
在 Twinkle 框架开源的同时，我们依托ModelScope的后台服务，也提供了托管的模型训练服务(Training as a Service），开发者可以通过这一服务，
免费体验Twinkle的训练API。

目前在集群中运行的模型是[Qwen/Qwen3-30B-A3B-Instruct-2507](https://www.modelscope.cn/models/Qwen/Qwen3-30B-A3B-Instruct-2507)。下面介绍具体的使用方法：

## Step 1. 注册魔搭用户
## Step 1. 注册ModelScope用户并申请加入 twinkle-explorers 组织

开发者首先需要注册成为魔搭用户，使用魔搭社区的 token 进行调用。
开发者首先需要注册成为ModelScope用户，并申请加入 [Twinkle-Explorers](https://modelscope.cn/organization/twinkle-explorers) 组织，
来获取访问权限。当前免费的Serverless训练体验，还在灰度测试中，暂时只向组织内的用户开放。您也可以通过本地部署服务，来使用Twinkle✨。

注册地址：https://www.modelscope.cn/

调用端点：https://www.modelscope.cn/twinkle
在注册并获批加入[Twinkle-Explorers](https://modelscope.cn/organization/twinkle-explorers) 组织后，在此页面获取
访问的API-Key（即ModelScope平台的访问Token）：https://www.modelscope.cn/my/access/token 。

token 在这里获取：https://www.modelscope.cn/my/access/token 拷贝访问令牌。
调用端点：`base_url="https://www.modelscope.cn/twinkle"`

## Step 2. 加入 twinkle-explorers 组织

目前 twinkle-kit 的远程训练能力在灰度测试中，开发者需要加入 [twinkle-explorers](https://www.modelscope.cn/models/twinkle-explorers) 组织，组织内的用户可以进行使用和测试。

## Step 3. 查看 Cookbook 并二次定制开发
## Step 2. 查看 Cookbook 并二次定制开发

我们强烈推荐开发者查看我们的 [cookbook](https://github.com/modelscope/twinkle/tree/main/cookbook/client/)，并根据其中的训练代码进行二次开发。

Expand Down
Loading