Downloads · 30 days
17
22% of all-time downloads
OpenSakura/OpenSakura-MODEL-260906-LN-ja-zh-SFT-qwen36-35b-a3b-sft
OpenSakura-MODEL-260906-LN-ja-zh-SFT-qwen36-35b-a3b-sft is a translation model from OpenSakura. Use it when you need text moved from one language to another. It is set up for transformers.
Japanese light-novel to Simplified Chinese translation SFT of the OpenSakura CPT base, exported to Hugging Face format.
Downloads · 30 days
17
22% of all-time downloads
All-time downloads
78
Public
Parameters
34.7B
69.3 GB on disk
Likes
0
Public
Click a slice to open those files.
.safetensors69.3 GB · 100%
From the Hugging Face model README
Japanese light-novel to Simplified Chinese translation SFT of the OpenSakura CPT base, exported to Hugging Face format.
runs/main/outputs/checkpoint-6003OpenSakura/OpenSakura-MODEL-260419-LN-ja-zh-PT-qwen36-35b-a3b-cpt543c3f95d2c08b43e3afcd1232e9b2859c0b264bmain_lilith65_eve350.65, Eve 0.35OpenSakura/OpenSakura-DS-260220-LN-ja-zh-COT-Lilith @ d6602ba7b2ff8fe95f73e1abc5ea83eb86173bd5OpenSakura/OpenSakura-DS-260220-LN-ja-zh-ALIGNED-Eve @ c0fbb13e06b1d95f456b6de7b942768641696543native_qwen_thinkingnative_qwen_empty_thinkingbfloat166003, 8x FSDP1, max_seq_length 122880.4496 (perplexity 1.568); train loss 0.28193.73e90.35, final loss weight 1.0The formatter builds a normal Qwen chat conversation with a system message and
a user message, then starts the assistant with Qwen native thinking tags.
你是专业的日文轻小说中文译者。请将输入的日文原文翻译为流畅、自然、具有轻小说/二次元语感的简体中文。
要求:
1. 严格保持行数一一对应:输入多少行,输出多少行,不合并、不拆分、不遗漏。
2. 根据上下文自然补全日文省略的主语,并保持角色口吻、性格、性别感和对话关系。
3. 保留原文的特殊格式,包括行首全角空格、转义符、特殊符号等。
4. Ruby/注音格式如“漢字(よみ)”若承载双关、设定、特殊称呼或魔法名,应在译文中保留“汉字(注音)”结构;普通读音标注则直接翻译含义。
5. 日文标点转换为中文标点,如「」转为“”,、转为,。
6. 忠实翻译原文的语气、粗俗度、暴力感、暧昧感和文学风格,不擅自删减、净化、美化或解释。
只输出译文纯文本,不输出说明、分析、JSON、开场白或结束语。
输入(总行数: {line_count}):
{source_text}
When glossary entries are present, the formatter prepends them before the input block:
术语表 <术语原文> -> <术语译文> #<术语信息>:
{source_term_1} -> {target_term_1} #{info_1}
{source_term_2} -> {target_term_2} #{info_2}
输入(总行数: {line_count}):
{source_text}
The tokenizer chat template renders the prompt as Qwen chat markers:
<|im_start|>system
{system_prompt}<|im_end|>
<|im_start|>user
{user_prompt}<|im_end|>
<|im_start|>assistant
<think>
For Eve empty-thinking examples, the assistant prefill is:
<think>
</think>
{final_translation}<|im_end|>
For Lilith thinking examples, the assistant target is:
<think>
{reasoning_content}
</think>
{final_translation}<|im_end|>
messages = [
{"role": "system", "content": SYSTEM_PROMPT},
{
"role": "user",
"content": "输入(总行数: 1):\nこれは例ではありません。",
},
]
text = tokenizer.apply_chat_template(
messages,
tokenize=False,
add_generation_prompt=True,
enable_thinking=True,
)
Note: the training user prompt header is Chinese 输入(总行数: N):.
Published as the current OpenSakura ja-zh SFT artifact. Downstream evaluation and preference work should initialize from this export.