魔搭社区免费 LLM 训练：Twinkle 训练即服务

import os
from tqdm import tqdm
from tinker import types
from twinkle_client import init_tinker_client
from twinkle.dataloader import DataLoader
from twinkle.dataset import Dataset, DatasetMeta
from twinkle.preprocessor import SelfCognitionProcessor
from twinkle.server.common import input_feature_to_datum

base_model = 'ms://Qwen/Qwen3-30B-A3B-Instruct-2507'
base_url = 'http://www.modelscope.cn/twinkle'
api_key = os.environ.get('MODELSCOPE_TOKEN')

# 加载并预处理数据集
dataset = Dataset(
    dataset_meta=DatasetMeta('ms://swift/self-cognition', data_slice=range(500))
)
dataset.set_template('Template', model_id=base_model, max_length=256)
dataset.map(
    SelfCognitionProcessor('Twinkle Model', 'ModelScope Team'), 
    load_from_cache_file=False
)
dataset.encode(batched=True, load_from_cache_file=False)
dataloader = DataLoader(dataset=dataset, batch_size=8)

# 初始化客户端
init_tinker_client()
from tinker import ServiceClient

service_client = ServiceClient(base_url=base_url, api_key=api_key)
training_client = service_client.create_lora_training_client(
    base_model=base_model[len('ms://'):], 
    rank=16
)

# 训练循环
for epoch in range(2):
    for step, batch in tqdm(enumerate(dataloader)):
        input_datum = [input_feature_to_datum(f) for f in batch]

        fwdbwd_future = training_client.forward_backward(input_datum, "cross_entropy")
        optim_future = training_client.optim_step(types.AdamParams(learning_rate=1e-4))

        fwdbwd_result = fwdbwd_future.result()
        optim_result = optim_future.result()
        print(f'Training Metrics: {optim_result}')

    # 保存检查点
    result = training_client.save_state(f"twinkle-lora-{epoch}").result()
    print(f'Saved checkpoint for epoch {epoch} to {result.path}')

import os
from tinker import types
from twinkle.data_format import Message, Trajectory
from twinkle.template import Template
from twinkle import init_tinker_client

init_tinker_client()
from tinker import ServiceClient

base_model = 'Qwen/Qwen3-30B-A3B-Instruct-2507'
base_url = 'http://www.modelscope.cn/twinkle'

service_client = ServiceClient(
    base_url=base_url,
    api_key=os.environ.get('MODELSCOPE_TOKEN')
)

# 加载训练好的 LoRA
sampling_client = service_client.create_sampling_client(
    model_path='twinkle://xxx-Qwen_Qwen3-30B-A3B-Instruct-2507-xxx/weights/twinkle-lora-1',
    base_model=base_model
)

# 准备提示词
template = Template(model_id=f'ms://{base_model}')
trajectory = Trajectory(
    messages=[
        Message(role='system', content='You are a helpful assistant'),
        Message(role='user', content='Who are you?'),
    ]
)

input_feature = template.encode(trajectory, add_generation_prompt=True)
prompt = types.ModelInput.from_ints(input_feature['input_ids'].tolist())

# 采样
params = types.SamplingParams(
    max_tokens=128,
    temperature=0.7,
    stop=['\n']
)

future = sampling_client.sample(prompt=prompt, sampling_params=params, num_samples=1)
result = future.result()

for i, seq in enumerate(result.sequences):
    print(f'{i}: {repr(template.decode(seq.tokens))}')