mmis1000/asmr-qwen3.5-9b-zh-cn-gguf-v0.1 overview

{
  "metadata": {},
  "card_data": {
    "license": "apache-2.0",
    "language": [
      "ja",
      "zh"
    ],
    "tags": [
      "asmr",
      "translation",
      "japanese",
      "chinese",
      "gguf"
    ],
    "base_model": "unsloth/Qwen3.5-9B",
    "quantized_by": "unsloth",
    "pipeline_tag": "text-generation",
    "frontmatter": {
      "license": "apache-2.0",
      "language": [
        "ja",
        "zh"
      ],
      "tags": [
        "asmr",
        "translation",
        "japanese",
        "chinese",
        "gguf"
      ],
      "base_model": "unsloth/Qwen3.5-9B",
      "quantized_by": "unsloth",
      "pipeline_tag": "text-generation"
    },
    "hero_image_url": "",
    "summary": "GGUF quantizations of a fine-tuned model for translating Japanese ASMR transcriptions (ASR/Whisper output) into **Simplified Chinese**. The model normalizes imperfect audio transcriptions, applies domain-specific glossaries, and translates character dialogue while retaining emotion and nuances.",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "---\nlicense: apache-2.0\nlanguage:\n  - ja\n  - zh\ntags:\n  - asmr\n  - translation\n  - japanese\n  - chinese\n  - gguf\nbase_model: unsloth/Qwen3.5-9B\nquantized_by: unsloth\npipeline_tag: text-generation\n---\n\n# asmr-qwen3.5-9b-zh-cn-gguf-v0.1\n\nGGUF quantizations of a fine-tuned model for translating Japanese ASMR transcriptions (ASR/Whisper output) into **Simplified Chinese**.\n\nThe model normalizes imperfect audio transcriptions, applies domain-specific glossaries, and translates character dialogue while retaining emotion and nuances.\n\n## Standard Mode\n\nThe traditional output format where only the translated text is returned.\n\n## Available Quantizations\n\n| Quantization | Filename | Size | Description |\n|---|---|---|---|\n| q4_k_m | `asmr-qwen3.5-9b-zh-cn-gguf-v0.1-q4_k_m.gguf` | 5.2 GB | Good balance of quality and size |\n| q6_k | `asmr-qwen3.5-9b-zh-cn-gguf-v0.1-q6_k.gguf` | 6.9 GB | Higher quality, moderate size |\n| q8_0 | `asmr-qwen3.5-9b-zh-cn-gguf-v0.1-q8_0.gguf` | 8.9 GB | Near-lossless quality |\n| bf16 | `asmr-qwen3.5-9b-zh-cn-gguf-v0.1-bf16.gguf` | 16.7 GB | Full BF16, no quantization loss |\n\n## Prompt Example\n\n````text\n将以下日语ASMR逐字稿翻译成简体中文。\n\n音轨：track01_示例音轨\n场景说明：主角与青梅竹马在校园下午的对话...\n\n术语表（请严格使用zh栏位的译名）：\n{\n  \"cvs\": [],\n  \"characters\": [],\n  \"terms\": [{\"ja\": \"放課後\", \"zh\": \"放学后\"}]\n}\n\n翻译前请静默修正以下Whisper识别错误：\n- 重复片语（连续3次以上且无变化）：仅保留一次\n- 错字／同音异字：依上下文修正\n- 字幕版权行（字幕：／翻訳：／QQ／LINE水印）：text设为null\n- 错误专有名词：依术语表修正\n\n翻译规则：\n- 呻吟与气息声（あ、ん、はあ）→ 自然对应（啊、嗯、哈、呼）\n- 拟声词：日语形式翻译（パンパン→啪啪）；中文形式保留原样\n- 保留角色语气与口吻\n- text字段只输出译文，不加注释或括号说明\n\n输入：逐字稿JSON数组 — {\"id\": <n>, \"text\": \"<日文>\", \"start\": <ms>, \"end\": <ms>}\n\n输出：将连续构成同一句话的片段合并，JSON数组格式：\n{\"ids\": [<n>, ...], \"text\": \"<简体中文>\", \"start\": <最早ms>, \"end\": <最晚ms>}\n\n字幕版权行：{\"ids\": [<n>], \"text\": null, \"start\": <ms>, \"end\": <ms>}\n每个输入id必须恰好出现在一个输出项中。\n\n逐字稿：\n[\n  {\"id\": 1, \"text\": \"ねぇ、放課後、\", \"start\": 3000, \"end\": 5000},\n  {\"id\": 2, \"text\": \"一緒に帰らない?\", \"start\": 5000, \"end\": 7000}\n]\n````\n\n**Example Output:**\n```json\n[{\"ids\": [1, 2], \"text\": \"呐，放学后，要不要一起回去？\", \"start\": 3000, \"end\": 7000}]\n```\n\n## Usage\n\n### llama-server\n\n```bash\nllama-server -m asmr-qwen3.5-9b-zh-cn-gguf-v0.1-q4_k_m.gguf -c 4096 --port 8080\n```\n\n### llama-cli\n\n```bash\nllama-cli -m asmr-qwen3.5-9b-zh-cn-gguf-v0.1-q4_k_m.gguf -p \"<your prompt>\" -n 2048\n```\n\n## Structured Decoding (Recommended)\n\nThis model outputs JSON arrays. Using structured decoding (e.g. GBNF grammar or JSON schema constraints) avoids wasted computation on malformed output and guarantees valid JSON on every generation.\n\n**JSON Schema:**\n```json\n{\n  \"type\": \"array\",\n  \"items\": {\n    \"type\": \"object\",\n    \"properties\": {\n      \"ids\": {\n        \"type\": \"array\",\n        \"items\": {\n          \"type\": \"integer\"\n        },\n        \"minItems\": 1\n      },\n      \"text\": {\n        \"anyOf\": [\n          {\n            \"type\": \"string\"\n          },\n          {\n            \"type\": \"null\"\n          }\n        ]\n      },\n      \"start\": {\n        \"type\": \"integer\"\n      },\n      \"end\": {\n        \"type\": \"integer\"\n      }\n    },\n    \"required\": [\n      \"ids\",\n      \"text\",\n      \"start\",\n      \"end\"\n    ],\n    \"additionalProperties\": false\n  },\n  \"minItems\": 1\n}\n```\n\nSupported by llama.cpp (`--json-schema`), vLLM, and `outlines`.\n\n## Training Details\n\n- **Base model**: `unsloth/Qwen3.5-9B`\n- **Method**: LoRA (r=16, alpha=16)\n- **Target modules**: q_proj, v_proj, up_proj, gate_proj, k_proj, down_proj, o_proj\n- **Locale**: zh-cn (Simplified Chinese)\n- **Mode**: Standard Mode\n- **Max sequence length**: 4096\n- **Precision**: bf16\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "asmr",
    "translation",
    "japanese",
    "chinese",
    "text-generation",
    "ja",
    "zh",
    "base_model:unsloth/Qwen3.5-9B",
    "base_model:quantized:unsloth/Qwen3.5-9B",
    "license:apache-2.0",
    "endpoints_compatible",
    "region:us",
    "conversational"
  ],
  "likes": 0,
  "downloads": 721,
  "gated": false,
  "private": false,
  "last_modified": "2026-04-09T07:55:07.000Z",
  "created_at": "2026-03-31T19:07:25.000Z",
  "pipeline_tag": "text-generation",
  "library_name": ""
}

{
  "_id": "69cc1b6da330c9c816669c31",
  "id": "mmis1000/asmr-qwen3.5-9b-zh-cn-gguf-v0.1",
  "modelId": "mmis1000/asmr-qwen3.5-9b-zh-cn-gguf-v0.1",
  "sha": "2e0bc0d8ef20fb6e52c251be44387a3f0e00e13c",
  "createdAt": "2026-03-31T19:07:25.000Z",
  "lastModified": "2026-04-09T07:55:07.000Z",
  "author": "mmis1000",
  "downloads": 721,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "text-generation",
  "library_name": "",
  "siblings_count": 6
}