[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-78045":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":8,"htmlUrl":8,"language":9,"languages":8,"totalLinesOfCode":8,"stars":10,"forks":11,"watchers":12,"openIssues":13,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":15,"stars7d":16,"stars30d":17,"stars90d":14,"forks30d":14,"starsTrendScore":18,"compositeScore":19,"rankGlobal":8,"rankLanguage":8,"license":20,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":23,"hasPages":21,"topics":24,"createdAt":8,"pushedAt":8,"updatedAt":25,"readmeContent":26,"aiSummary":27,"trendingCount":14,"starSnapshotCount":14,"syncStatus":28,"lastSyncTime":29,"discoverSource":30},78045,"Hy-MT2","Tencent-Hunyuan\u002FHy-MT2","Tencent-Hunyuan",null,"Python",426,32,3,5,0,1,13,361,7,60.06,"Other",false,"main",true,[],"2026-06-12 04:01:23","\u003Cp align=\"left\">\n   English ｜ \u003Ca href=\"README_CN.md\">中文\u003C\u002Fa>\n\u003C\u002Fp>\n\u003Cbr>\n\n\u003Cp align=\"center\">\n \u003Cimg src=\"imgs\u002Flogo-en.png\" width=\"400\"\u002F> \u003Cbr> \n\u003C\u002Fp>\n\n\u003Cdiv align=\"center\" style=\"line-height: 1;\">\n\n\n[![HuggingFace](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A4%97%20Hugging%20Face-Tencent%20Hy-ffc107?color=ffc107&logoColor=white)](https:\u002F\u002Fhuggingface.co\u002Fcollections\u002Ftencent\u002Fhy-mt2)\n  \n[![ModelScope](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FModelScope-Tencent%20Hy-624aff)](https:\u002F\u002Fmodelscope.cn\u002Fcollections\u002FTencent-Hunyuan\u002FHy-MT2)\n\n\u003C\u002Fdiv>\n\n\u003Cp align=\"center\">\n    🖥️ \u003Ca href=\"https:\u002F\u002Faistudio.tencent.com\u002Fllm\u002Fen?tabIndex=0\">\u003Cb>Official Website\u003C\u002Fb>\u003C\u002Fa>  |  \n    💬 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FTencent-Hunyuan\u002FHy-MT2\">\u003Cb>GitHub\u003C\u002Fb>\u003C\u002Fa>  |  \n    🪡 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002FTencent\u002FAngelSlim\u002Ftree\u002Fmain\">\u003Cb>AngelSlim\u003C\u002Fb>\u003C\u002Fa>  |  \n    📚 \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fpdf\u002F2605.22064\">\u003Cb>Hy-MT2 Report\u003C\u002Fb>\u003C\u002Fa>\u003C\u002Fp>\n\n## Model Introduction\n\n\nHy-MT2 is a family of “fast-thinking” multilingual translation models designed for complex real-world scenarios. It includes three model sizes: 1.8B, 7B, and 30B-A3B (MoE), all of which support translation among 33 languages and effectively follow translation instructions in multiple languages.\nFor on-device deployment, AngelSlim 1.25-bit extreme quantization reduces the storage requirement of the 1.8B model to only 440 MB and improves inference speed by 1.5x.\nMulti-dimensional evaluations show that Hy-MT2 delivers outstanding performance across general, real-world business, domain-specific, and instruction-following translation tasks. The 7B and 30B-A3B models outperform open-source models such as DeepSeek-V4-Pro and Kimi K2.6 in fast-thinking mode, while the lightweight 1.8B model also surpasses mainstream commercial APIs from providers such as Microsoft and Doubao overall.\n\nIn this release, we also open-source [IFMTBench](.\u002FIFMTBench\u002FREADME.md), a benchmark for evaluating translation instruction-following capabilities.\n\nWe also welcome everyone to use our released Hy-MT2-Translator Skill, which makes it easy to integrate Hy-MT2 series models for translation tasks. Download links: [ClawHub](https:\u002F\u002Fclawhub.ai\u002Ftencent-adm\u002Fhy-mt2-translator-skill) and [SkillHub](https:\u002F\u002Fskillhub.cn\u002Fskills\u002Fhy-mt2-translator).\n\nNow, Tencent Hy is officially partnering with WMT26 for the \"Video Subtitle Translation Task\" (https:\u002F\u002Fwww2.statmt.org\u002Fwmt26\u002Fvideo-subtitle-translation.html). Participants who use the Hy-MT model series to compete in the \"General Machine Translation Task\" (https:\u002F\u002Fwww2.statmt.org\u002Fwmt26\u002Ftranslation-task.html) and the \"Video Subtitle Translation Task\" will have the chance to win special awards sponsored by Hunyuan. We sincerely invite everyone to participate and jointly push the boundaries of machine translation technology!\n\n## News\n\n* 2026.5.21  We open-sourced **Hy-MT2-1.8B**\u002F**Hy-MT2-7B**\u002F**Hy-MT2-30B-A3B**\u002F**IFMTBench** on HuggingFace and ModelScope.\n* 2025.12.30 We open-sourced **HY-MT1.5-1.8B** and **HY-MT1.5-7B** on HuggingFace and ModelScope.\n* 2025.9.1 We open-sourced **Hunyuan-MT-7B** and **Hunyuan-MT-Chimera-7B** on HuggingFace and ModelScope.\n\n\n## Results\n\u003Cdiv align='center'>\n\u003Cimg src=\"imgs\u002Fmain_result.png\" width = \"100%\" \u002F>\n\u003C\u002Fdiv>\n\nFor more experimental results and analysis, please refer to our [report](https:\u002F\u002Farxiv.org\u002Fpdf\u002F2605.22064).\n\n \n\n## Model Links\n| Model Name  | Description | Download Link |\n| ----------- | ----------- |-----------\n| Hy-MT2-1.8B  | Hy 1.8B translation model |🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-1.8B)|\n| Hy-MT2-1.8B-FP8 | Hy 1.8B translation model, FP8 quantization    | 🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-1.8B-FP8)|\n| Hy-MT2-1.8B-GGUF | Hy 1.8B translation model, llama.cpp    | 🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-1.8B-GGUF)|\n| Hy-MT2-1.8B-2bit-GGUF | Hy 1.8B translation model, llama.cpp, 2bit    | 🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-1.8B-2bit-GGUF)|\n| Hy-MT2-1.8B-1.25bit-GGUF | Hy 1.8B translation model, llama.cpp, 1.25bit    | 🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-1.8B-1.25bit-GGUF)|\n| Hy-MT2-7B | Hy 7B translation model    | 🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-7B)|\n| Hy-MT2-7B-FP8 | Hy 7B translation model, FP8 quantization     | 🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-7B-FP8)|\n| Hy-MT2-7B-GGUF | Hy 7B translation model, llama.cpp    | 🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-7B-GGUF)|\n| Hy-MT2-30B-A3B | Hy 30B-A3B translation model    | 🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-30B-A3B)|\n| Hy-MT2-30B-A3B-FP8 | Hy 30B-A3B translation model, FP8 quantization     | 🤗 [Model](https:\u002F\u002Fhuggingface.co\u002Ftencent\u002FHy-MT2-30B-A3B-FP8)|\n\n\n\n## Hy-MT2 Translation Task Instruction Examples (Chinese-English Comparison)\n\n*Note: In the following examples, both source_lang and target_lang should use the full language names. Chinese names should be used in Chinese prompts, and English names should be used in English prompts.*\n\n| Type | Chinese prompt | English prompt |\n|---|---|---|\n| **Default Translation** | 将以下文本翻译为 `{target_lang}`，注意**只需要输出翻译后的结果，不要额外解释**：\u003Cbr>\u003Cbr>`{source_text}` | Translate the following text into `{target_lang}`. Note that you should **only output the translated result without any additional explanation**:\u003Cbr>\u003Cbr>`{source_text}` |\n| **Terminology** | *参考下面的翻译：*\u003Cbr>`{text}` 翻译成 `{text}`\u003Cbr>`{text}` 翻译成 `{text}`\u003Cbr>`{text}` 翻译成 `{text}`\u003Cbr>将以下文本翻译为 `{target_lang}`，注意**只需要输出翻译后的结果，不要额外解释**：\u003Cbr>\u003Cbr>`{source_text}` | *Reference the following translations:*\u003Cbr>`{text}` translates to `{text}`\u003Cbr>`{text}` translates to `{text}`\u003Cbr>`{text}` translates to `{text}`\u003Cbr>\u003Cbr>Translate the following text into `{target_lang}`. Note that you must **ONLY output the translated result without any additional explanation**:\u003Cbr>\u003Cbr>`{source_text}` |\n| **Style** | 请将以下文本翻译为 `{target_lang}`。\u003Cbr>注意翻译的风格要严格符合【**`{target_style}`**】\u003Cbr>\u003Cbr>`{source_text}` | Please translate the following text into `{target_lang}`. Note that the translation style must strictly conform to [**`{target_style}`**]:\u003Cbr>\u003Cbr>`{source_text}` |\n| **Personalization** | *【待翻译文本】*\u003Cbr>`{source_text}`\u003Cbr>\u003Cbr>*【翻译任务】*\u003Cbr>1、**`{user_preferences}`**\u003Cbr>2、**`{user_preferences}`**\u003Cbr>3、……\u003Cbr>4、将【待翻译文本】翻译为 `{target_lang}`。 | *[Source Text]*\u003Cbr>`{source_text}`\u003Cbr>\u003Cbr>*[Translation Tasks]*\u003Cbr>1. **`{user_preferences}`**\u003Cbr>2. **`{user_preferences}`**\u003Cbr>3. ...\u003Cbr>4. Translate the [Source Text] into `{target_lang}`. |\n| **Delimiters** | 请将以下文本准确翻译为 `{target_lang}`。\u003Cbr>你必须在译文中**保留等量的分隔符，绝对不可遗漏、转义或翻译该符号，并注意分隔符的位置**。\u003Cbr>\u003Cbr>`{source_text}` | Please accurately translate the following text into `{target_lang}`.\u003Cbr>You must **retain the exact same number of delimiters in the translation. Strictly do not omit, escape, or translate these symbols, and pay close attention to their placement**.\u003Cbr>\u003Cbr>`{source_text}` |\n| **Structured Data 1** | *# 任务目标*\u003Cbr>将下方 `{source_text}` 中的 `{format_type}` 格式数据翻译为 `{target_lang}`。\u003Cbr>\u003Cbr>*# 严格约束*\u003Cbr>1. **结构锁定**：绝对保持原有的 `{format_type}` 数据结构、缩进和层级完全不变。\u003Cbr>2. **选择性翻译**：仅翻译面向用户展示的可见文本内容。\u003Cbr>3. **禁止修改**：**严禁**翻译或更改任何代码标签、键名 (Key)、变量占位符（如 `{{var}}`、`${var}`、`%s`、`%d` 等）或代码属性。\u003Cbr>\u003Cbr>*# 数据输入*\u003Cbr>`{source_text}` | *### Task*\u003Cbr>Translate the user-facing text within the following `{format_type}` data into `{target_lang}`.\u003Cbr>\u003Cbr>*### Strict Rules*\u003Cbr>1. **Structure Preservation:** You MUST preserve the original `{format_type}` data structure, nesting, hierarchy, and indentation exactly as they are.\u003Cbr>2. **Selective Translation:** Translate ONLY the visible, user-facing text content\u002Fvalues.\u003Cbr>3. **Strict Non-Translation:** NEVER translate or alter code tags, keys, properties, object names, or variable placeholders. Leave them exactly in their original English\u002Fcode form.\u003Cbr>\u003Cbr>*### Source Data*\u003Cbr>`{source_text}` |\n| **Structured Data 2** | *【背景信息】*\u003Cbr>`{background_text}`\u003Cbr>\u003Cbr>请结合背景信息将以下文本翻译为 `{target_lang}`。\u003Cbr>\u003Cbr>*【待翻译文本】*\u003Cbr>`{source_text}` | *[Background Information]*\u003Cbr>`{background_text}`\u003Cbr>\u003Cbr>Please translate the following text into `{target_lang}`, taking the provided background information into consideration.\u003Cbr>\u003Cbr>*[Source Text]*\u003Cbr>`{source_text}` |\n\n---\n\n## Inference and Deployment\n\nFor 1.8B and 7B, we recommend using the following parameters for inference. Note that our models do not have a default system_prompt.\n\n```json\n\n{\n  \"temperature\": 0.7,\n  \"top_p\": 0.6,\n  \"top_k\": 20,\n  \"repetition_penalty\": 1.05,\n  \"max_tokens\": 4096\n}\n```\n\nFor 30B-A3B, we recommend using the following parameters for inference. Note that our models do not have a default system_prompt.\n\n```json\n\n{\n  \"temperature\": 0.7,\n  \"top_p\": 1.0,\n  \"top_k\": -1,\n  \"repetition_penalty\": 1.0,\n  \"max_tokens\": 4096\n}\n```\n\n\n### transformers\n\ntransformers>=5.6.0\n\n```python\nfrom transformers import AutoModelForCausalLM, AutoTokenizer\nimport torch\n\nmodel_path = \"tencent\u002FHy-MT2-30B-A3B\"\n\n# Load tokenizer\ntokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)\n\n# Load model\nmodel = AutoModelForCausalLM.from_pretrained(\n    model_path,\n    dtype=torch.bfloat16,\n    device_map=\"auto\",\n    trust_remote_code=True,\n)\n\nmodel.eval()\n\n# Example inference\nprompt = \"将以下文本翻译成英语,注意只需要输出翻译后的结果,不要额外解释:\\n\\n今天天气真好。\"\nmessages = [{\"role\": \"user\", \"content\": prompt}]\ninputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors=\"pt\").to(model.device)\n\nwith torch.no_grad():\n    outputs = model.generate(\n        **inputs,\n        max_new_tokens=4096,\n    )\nresponse = tokenizer.decode(outputs[0][inputs[\"input_ids\"].shape[-1]:], skip_special_tokens=True)\nprint(response)\n```\n\n### vllm\n\nBuild vLLM from source:\n```bash\nuv venv --python 3.12 --seed --managed-python\nsource .venv\u002Fbin\u002Factivate\ngit clone https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm.git\ncd vllm\nuv pip install --editable . --torch-backend=auto\n```\n\nStart the vLLM server:\n\n```bash\nvllm serve tencent\u002FHy-MT2-30B-A3B --tensor-parallel-size 1\n```\n\n### sglang\n\nBuild SGLang from source:\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fsgl-project\u002Fsglang\ncd sglang\npip3 install pip --upgrade\npip3 install \"transformers>=5.6.0\"\npip3 install -e \"python\"\n```\n\nLaunch SGLang server:\n\n```bash\npython3 -m sglang.launch_server --model tencent\u002FHy-MT2-30B-A3B --tp 1\n```\n\n### llama_cpp\n**❕❕ This gguf depends on our STQ kernel, which is released at [PR #22836](https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp\u002Fpull\u002F22836).**\n\n#### Clone llama.cpp\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Fggml-org\u002Fllama.cpp.git\n```\n\n#### Enter the llama.cpp folder\n\n```bash\ncd llama.cpp\n```\n\n#### Build llama.cpp\n\n```bash\ncmake -B build\ncmake --build build --config Release\n```\n\n#### Run a completion example\n\n```bash\n.\u002Fbuild\u002Fbin\u002Fllama-completion \\\n  --model model.gguf  \\\n  -p \"Translate the following segment into Chinese, without additional explanation：Hello\" \\\n  --jinja \\\n  -ngl 0 \\\n  -n 64 -st \n```\n\n#### Run the llama.cpp benchmark\n\n```bash\n.\u002Fbuild\u002Fbin\u002Fllama-bench -m model_zoo\u002Fmodel.gguf  -ngl 0\n```\n\n\n## Model Training\nHy-MT2 provides a complete model training pipeline, supporting both full-parameter fine-tuning and LoRA fine-tuning, as well as multiple DeepSpeed ZeRO configurations and LLaMA-Factory integration.\n\nFor detailed training documentation, please refer to: [Model Training Guide](.\u002Ftrain\u002FREADME.md)\n\n## Quantization Tool\n\nWe provide [AngelSlim](https:\u002F\u002Fgithub.com\u002Ftencent\u002FAngelSlim), an easy-to-use, comprehensive, and efficient large model compression toolkit covering common quantization algorithms, low-bit quantization, speculative sampling, and more.\n\n\n## Supported Languages\n| Languages         | Abbr.   | Chinese Names   |\n|-------------------|---------|-----------------|\n| Chinese           | zh      | 中文            |\n| English           | en      | 英语            |\n| French            | fr      | 法语            |\n| Portuguese        | pt      | 葡萄牙语        |\n| Spanish           | es      | 西班牙语        |\n| Japanese          | ja      | 日语            |\n| Turkish           | tr      | 土耳其语        |\n| Russian           | ru      | 俄语            |\n| Arabic            | ar      | 阿拉伯语        |\n| Korean            | ko      | 韩语            |\n| Thai              | th      | 泰语            |\n| Italian           | it      | 意大利语        |\n| German            | de      | 德语            |\n| Vietnamese        | vi      | 越南语          |\n| Malay             | ms      | 马来语          |\n| Indonesian        | id      | 印尼语          |\n| Filipino          | tl      | 菲律宾语        |\n| Hindi             | hi      | 印地语          |\n| Traditional Chinese | zh-Hant| 繁体中文        |\n| Polish            | pl      | 波兰语          |\n| Czech             | cs      | 捷克语          |\n| Dutch             | nl      | 荷兰语          |\n| Khmer             | km      | 高棉语          |\n| Burmese           | my      | 缅甸语          |\n| Persian           | fa      | 波斯语          |\n| Gujarati          | gu      | 古吉拉特语      |\n| Urdu              | ur      | 乌尔都语        |\n| Telugu            | te      | 泰卢固语        |\n| Marathi           | mr      | 马拉地语        |\n| Hebrew            | he      | 希伯来语        |\n| Bengali           | bn      | 孟加拉语        |\n| Tamil             | ta      | 泰米尔语        |\n| Ukrainian         | uk      | 乌克兰语        |\n| Tibetan           | bo      | 藏语            |\n| Kazakh            | kk      | 哈萨克语        |\n| Mongolian         | mn      | 蒙古语          |\n| Uyghur            | ug      | 维吾尔语        |\n| Cantonese         | yue     | 粤语            |\n\n\n\n## Citing Hy-MT2\n\n```bibtex\n@misc{zheng2026hymt2familyfastefficient,\n      title={Hy-MT2: A Family of Fast, Efficient and Powerful Multilingual Translation Models in the Wild}, \n      author={Mao Zheng and Zheng Li and Tao Chen and Bo Lv and Mingrui Sun and Mingyang Song and Jinlong Song and Hong Huang and Decheng Wu and Hai Wang and Yifan Song and Yanfeng Chen and Guanwei Zhang},\n      year={2026},\n      eprint={2605.22064},\n      archivePrefix={arXiv},\n      primaryClass={cs.CL},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2605.22064}, \n} \n```\n\n## Contact Us\nIf you would like to leave feedback for our R&D and product teams, you are welcome to contact the Tencent Hunyuan LLM team. You can reach us by email at hunyuan_opensource@tencent.com.\n","Hy-MT2 是一系列为复杂现实场景设计的多语言翻译模型，支持33种语言之间的互译并能有效遵循多种语言的翻译指令。该项目的核心功能包括不同规模（1.8B、7B和30B-A3B）的模型选择，其中1.8B版本通过AngelSlim 1.25位极量量化技术将存储需求降至440MB，并提升了1.5倍的推理速度。Hy-MT2在通用、商业、特定领域及指令跟随翻译任务中表现出色，尤其是7B和30B-A3B型号超越了如DeepSeek-V4-Pro等开源模型。此外，项目还开源了一个名为IFMTBench的基准测试工具，用于评估模型对翻译指令的执行能力。Hy-MT2适用于需要高效准确跨语言交流的各种场景，特别是那些对设备资源有限制但要求高性能的应用场合。",2,"2026-06-11 03:56:23","CREATED_QUERY"]