Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
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
2 changes: 1 addition & 1 deletion content/_index.md
Original file line number Diff line number Diff line change
Expand Up @@ -138,7 +138,7 @@ sections:
# Use ModelScope's official TaaS endpoint — free, no local GPU needed
base_url = 'https://www.modelscope.cn/twinkle'
api_key = os.environ.get('MODELSCOPE_TOKEN')
base_model = 'Qwen/Qwen3.6-27B'
base_model = 'Qwen/Qwen3.8-27B'

# Prepare data locally
dataset = Dataset(dataset_meta=DatasetMeta('ms://swift/self-cognition'))
Expand Down
2 changes: 1 addition & 1 deletion content/_index.zh.md
Original file line number Diff line number Diff line change
Expand Up @@ -138,7 +138,7 @@ sections:
# 使用魔搭社区官方 TaaS 端点 — 免费,无需本地 GPU
base_url = 'https://www.modelscope.cn/twinkle'
api_key = os.environ.get('MODELSCOPE_TOKEN')
base_model = 'Qwen/Qwen3.6-27B'
base_model = 'Qwen/Qwen3.8-27B'

# 本地准备数据
dataset = Dataset(dataset_meta=DatasetMeta('ms://swift/self-cognition'))
Expand Down
14 changes: 7 additions & 7 deletions content/blog/modelscope-taas/index.md
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@ We're excited to announce that **Twinkle Training-as-a-Service (TaaS)** is now a

Training-as-a-Service lets you fine-tune large language models through a simple API, without managing infrastructure. The model runs on ModelScope's backend servers; you just send data and receive trained adapters.

Currently available model: **[Qwen/Qwen3.6-27B](https://www.modelscope.cn/models/Qwen/Qwen3.6-27B)**
Currently available model: **[Qwen/Qwen3.8-27B](https://www.modelscope.cn/models/Qwen/Qwen3.8-27B)**

## Getting Started

Expand All @@ -44,8 +44,8 @@ from twinkle.dataset import Dataset, DatasetMeta
from twinkle.preprocessor import SelfCognitionProcessor
from twinkle.server.common import input_feature_to_datum

base_model = 'Qwen/Qwen3.6-27B'
base_url = 'http://www.modelscope.cn/twinkle'
base_model = 'Qwen/Qwen3.8-27B'
base_url = 'https://www.modelscope.cn/twinkle'
api_key = os.environ.get('MODELSCOPE_TOKEN')

# Load and preprocess dataset
Expand Down Expand Up @@ -101,8 +101,8 @@ from twinkle import init_tinker_client
init_tinker_client()
from tinker import ServiceClient

base_model = 'Qwen/Qwen3.6-27B'
base_url = 'http://www.modelscope.cn/twinkle'
base_model = 'Qwen/Qwen3.8-27B'
base_url = 'https://www.modelscope.cn/twinkle'

service_client = ServiceClient(
base_url=base_url,
Expand All @@ -111,7 +111,7 @@ service_client = ServiceClient(

# Load your trained LoRA
sampling_client = service_client.create_sampling_client(
model_path='twinkle://xxx-Qwen_Qwen3.6-27B-xxx/weights/twinkle-lora-1',
model_path='twinkle://xxx-Qwen_Qwen3.8-27B-xxx/weights/twinkle-lora-1',
base_model=base_model
)

Expand Down Expand Up @@ -160,7 +160,7 @@ The free tier has some limitations:
|-----------|-------|
| Max Rank | 32 |
| modules_to_save | Not supported |
| Multimodal | Text-only (Qwen3.6-27B) |
| Multimodal | Text-only (Qwen3.8-27B) |

## What You Can Customize

Expand Down
14 changes: 7 additions & 7 deletions content/blog/modelscope-taas/index.zh.md
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@ categories:

训练即服务(Training-as-a-Service)让你通过简单的 API 微调大语言模型,无需管理基础设施。模型运行在魔搭的后台服务器上;你只需发送数据,即可获得训练好的适配器。

当前可用模型:**[Qwen/Qwen3.6-27B](https://www.modelscope.cn/models/Qwen/Qwen3.6-27B)**
当前可用模型:**[Qwen/Qwen3.8-27B](https://www.modelscope.cn/models/Qwen/Qwen3.8-27B)**

## 快速开始

Expand All @@ -44,8 +44,8 @@ from twinkle.dataset import Dataset, DatasetMeta
from twinkle.preprocessor import SelfCognitionProcessor
from twinkle.server.common import input_feature_to_datum

base_model = 'Qwen/Qwen3.6-27B'
base_url = 'http://www.modelscope.cn/twinkle'
base_model = 'Qwen/Qwen3.8-27B'
base_url = 'https://www.modelscope.cn/twinkle'
api_key = os.environ.get('MODELSCOPE_TOKEN')

# 加载并预处理数据集
Expand Down Expand Up @@ -101,8 +101,8 @@ from twinkle import init_tinker_client
init_tinker_client()
from tinker import ServiceClient

base_model = 'Qwen/Qwen3.6-27B'
base_url = 'http://www.modelscope.cn/twinkle'
base_model = 'Qwen/Qwen3.8-27B'
base_url = 'https://www.modelscope.cn/twinkle'

service_client = ServiceClient(
base_url=base_url,
Expand All @@ -111,7 +111,7 @@ service_client = ServiceClient(

# 加载训练好的 LoRA
sampling_client = service_client.create_sampling_client(
model_path='twinkle://xxx-Qwen_Qwen3.6-27B-xxx/weights/twinkle-lora-1',
model_path='twinkle://xxx-Qwen_Qwen3.8-27B-xxx/weights/twinkle-lora-1',
base_model=base_model
)

Expand Down Expand Up @@ -160,7 +160,7 @@ for i, seq in enumerate(result.sequences):
|------|------|
| 最大 Rank | 32 |
| modules_to_save | 不支持 |
| 多模态 | 仅支持文本(Qwen3.6-27B) |
| 多模态 | 仅支持文本(Qwen3.8-27B) |

## 可自定义的内容

Expand Down
5 changes: 3 additions & 2 deletions content/docs/guide/taas.md
Original file line number Diff line number Diff line change
Expand Up @@ -23,8 +23,8 @@ from twinkle.dataset import Dataset, DatasetMeta
from twinkle.preprocessor import SelfCognitionProcessor
from twinkle.server.common import input_feature_to_datum

# The base model (currently Qwen3.6-27B)
base_model = 'Qwen/Qwen3.6-27B'
# The base model (currently Qwen3.8-27B)
base_model = 'Qwen/Qwen3.8-27B'

# Prepare dataset
dataset = Dataset(dataset_meta=DatasetMeta(
Expand Down Expand Up @@ -167,6 +167,7 @@ client = init_twinkle_client(

| Model | Size | HuggingFace ID | Megatron |
|:------|:-----|:---------------|:---------|
| Qwen3.8 | 27B | Qwen/Qwen3.8-* | Yes |
| Qwen3.6 | 4B-35B-A3B | Qwen/Qwen3.6-* | Yes |
| Qwen3.5 | 2B-27B | Qwen/Qwen3.5-* | Yes |
| Qwen3 | 0.6B-32B | Qwen/Qwen3-* | Yes |
Expand Down
5 changes: 3 additions & 2 deletions content/docs/guide/taas.zh.md
Original file line number Diff line number Diff line change
Expand Up @@ -23,8 +23,8 @@ from twinkle.dataset import Dataset, DatasetMeta
from twinkle.preprocessor import SelfCognitionProcessor
from twinkle.server.common import input_feature_to_datum

# 基座模型(当前为 Qwen3.6-27B)
base_model = 'Qwen/Qwen3.6-27B'
# 基座模型(当前为 Qwen3.8-27B)
base_model = 'Qwen/Qwen3.8-27B'

# 准备数据集
dataset = Dataset(dataset_meta=DatasetMeta(
Expand Down Expand Up @@ -167,6 +167,7 @@ client = init_twinkle_client(

| 模型 | 规模 | HuggingFace ID | Megatron |
|:-----|:-----|:---------------|:---------|
| Qwen3.8 | 27B | Qwen/Qwen3.8-* | 支持 |
| Qwen3.6 | 4B-35B-A3B | Qwen/Qwen3.6-* | 支持 |
| Qwen3.5 | 2B-27B | Qwen/Qwen3.5-* | 支持 |
| Qwen3 | 0.6B-32B | Qwen/Qwen3-* | 支持 |
Expand Down
Loading
Loading