richarderkhov/vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf overview

{
  "metadata": {},
  "card_data": {
    "frontmatter": {},
    "hero_image_url": "",
    "summary": "Quantization made by Richard Erkhov. Github Discord Request more models salt-asr_wav-uni_1_tts_wav-uni_1-12k - GGUF | Name | Quant method | Size | | ---- | ---- | ---- | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q2_K.gguf | Q2_K | 1.28GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ3_XS.gguf | IQ3_XS | 1.39GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ3_S.gguf | IQ3_S | 1.45GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K_S.gguf | Q3_K_S | 1.45GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ3_M.gguf | IQ3_M | 1.5GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K.gguf | Q3_K | 1.58GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K_M.gguf | Q3_K_M | 1.58GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K_L.gguf | Q3_K_L | 1.7GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ4_XS.gguf | IQ4_XS | 1.72GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_0.gguf | Q4_0 | 1.8GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ4_NL.gguf | IQ4_NL | 1.8GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_K_S.gguf | Q4_K_S | 1.81GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_K.gguf | Q4_K | 1.89GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_K_M.gguf | Q4_K_M | 1.89GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_1.gguf | Q4_1 | 1.96GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_0.gguf | Q5_0 | 2.12GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_K_S.gguf | Q5_K_S | 2.12GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_K.gguf | Q5_K | 2.17GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_K_M.gguf | Q5_K_M | 2.17GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_1.gguf | Q5_1 | 2.29GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q6_K.gguf | Q6_K | 2.47GB | | salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q8_0.gguf | Q8_0 | 3.2GB | Original model description: ### English Version 🇬🇧 --- #### **Model Performance Overview** **Metrics**: | Model                     | PESQ@200       | STOI@200      | SI-SDR@200        | SIM-O@200      | |---------------------------|----------------|---------------|-------------------|----------------| | Original (LibriSpeech)    | 4.15           | 0.997         | 27.45 ±1.09       | —              | | Parler TTS Mini v1        | 1.29 ±0.49     | 0.15 ±0.12    | 25.0 ±2.9         | 0.88 ±0.03     | | Fish Speech 1.5           | 1.26 ±0.38     | 0.17 ±0.12    | 25.0 ±3.2         | 0.91 ±0.02     | | **Salt-ASR Wav-Uni 1-12k **  | **1.27 ±0.40**     | 0.18 ±0.09    | 20.3 ±3.69        | 0.88 ±0.02     | --- #### **Our Solution** --- #### **Resources** --- ### Русская Версия 🇷🇺 --- #### **Сравнение моделей** **Метрики**: | Модель                   | PESQ@200       | STOI@200      | SI-SDR@200        | SIM-O@200      | |--------------------------|----------------|---------------|-------------------|----------------| | Original (LibriSpeech)   | 4.15           | 0.997         | 27.45 ±1.09       | —              | | Parler TTS Mini v1       | 1.25 ±0.49     | 0.15 ±0.12    | 25.0 ±2.9         | 0.88 ±0.03     | | Fish Speech 1.5          | 1.26 ±0.38     | 0.17 ±0.12    | 25.0 ±3.2         | 0.91 ±0.02     | | **Salt-ASR Wav-Uni 1-12k **  | **1.27 ±0.40**     | 0.18 ±0.09    | 20.3 ±3.69        | 0.88 ±0.02     | --- #### **Наше решение** --- #### **Ресурсы** --- **Примечание**: Модель поддерживает генерацию коротких фраз на английском, немецком и французском.",
    "quick_links": [],
    "benchmark_table_html": "",
    "readme_markdown": "Quantization made by Richard Erkhov.\n\n[Github](https://github.com/RichardErkhov)\n\n[Discord](https://discord.gg/pvy7H8DZMG)\n\n[Request more models](https://github.com/RichardErkhov/quant_request)\n\n\nsalt-asr_wav-uni_1_tts_wav-uni_1-12k - GGUF\n- Model creator: https://huggingface.co/Vikhrmodels/\n- Original model: https://huggingface.co/Vikhrmodels/salt-asr_wav-uni_1_tts_wav-uni_1-12k/\n\n\n| Name | Quant method | Size |\n| ---- | ---- | ---- |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q2_K.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q2_K.gguf) | Q2_K | 1.28GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ3_XS.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ3_XS.gguf) | IQ3_XS | 1.39GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ3_S.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ3_S.gguf) | IQ3_S | 1.45GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K_S.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K_S.gguf) | Q3_K_S | 1.45GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ3_M.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ3_M.gguf) | IQ3_M | 1.5GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K.gguf) | Q3_K | 1.58GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K_M.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K_M.gguf) | Q3_K_M | 1.58GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K_L.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q3_K_L.gguf) | Q3_K_L | 1.7GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ4_XS.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ4_XS.gguf) | IQ4_XS | 1.72GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_0.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_0.gguf) | Q4_0 | 1.8GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ4_NL.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.IQ4_NL.gguf) | IQ4_NL | 1.8GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_K_S.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_K_S.gguf) | Q4_K_S | 1.81GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_K.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_K.gguf) | Q4_K | 1.89GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_K_M.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_K_M.gguf) | Q4_K_M | 1.89GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_1.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q4_1.gguf) | Q4_1 | 1.96GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_0.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_0.gguf) | Q5_0 | 2.12GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_K_S.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_K_S.gguf) | Q5_K_S | 2.12GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_K.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_K.gguf) | Q5_K | 2.17GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_K_M.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_K_M.gguf) | Q5_K_M | 2.17GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_1.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q5_1.gguf) | Q5_1 | 2.29GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q6_K.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q6_K.gguf) | Q6_K | 2.47GB |\n| [salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q8_0.gguf](https://huggingface.co/RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf/blob/main/salt-asr_wav-uni_1_tts_wav-uni_1-12k.Q8_0.gguf) | Q8_0 | 3.2GB |\n\n\n\n\nOriginal model description:\n### English Version 🇬🇧\n\n---\n\n#### **Model Performance Overview**  \n**Metrics**:  \n- **PESQ@200**: Perceptual Evaluation of Speech Quality (higher = better).  \n- **STOI@200**: Short-Time Objective Intelligibility (closer to 1 = better).  \n- **SI-SDR@200**: Scale-Invariant Signal-to-Distortion Ratio (higher = better).  \n- **SIM-O@200**: Similarity to ground truth (higher = better).  \n\n| Model                     | PESQ@200       | STOI@200      | SI-SDR@200        | SIM-O@200      |  \n|---------------------------|----------------|---------------|-------------------|----------------|  \n| Original (LibriSpeech)    | 4.15           | 0.997         | 27.45 ±1.09       | —              |  \n| Parler TTS Mini v1        | 1.29 ±0.49     | 0.15 ±0.12    | 25.0 ±2.9         | 0.88 ±0.03     |  \n| Fish Speech 1.5           | 1.26 ±0.38     | 0.17 ±0.12    | 25.0 ±3.2         | 0.91 ±0.02     |  \n| **Salt-ASR Wav-Uni 1-12k **  | **1.27 ±0.40**     | 0.18 ±0.09    | 20.3 ±3.69        | 0.88 ±0.02     |  \n\n---\n\n#### **Our Solution**  \n- **Method**: Extends a pre-trained LLM with audio tokens and fine-tunes on **TTS** and **ASR** tasks.  \n- **Training**:  \n  - SpeechTokenizer (semantic + audio tokens) outperformed Encodec (loss explosions resolved with TF32 precision).  \n  - Training time: **150 A100 GPU hours**.  \n- **Advantages**: Unified LM loss for dual tasks, minimal training overhead.  \n\n\n---\n\n#### **Resources**  \n- Code: [GitHub Repo](https://github.com/VikhrModels/Vikhr4o)  \n- Inference Demo: [Google Colab](https://colab.research.google.com/drive/1Poz6jNJu7-HRIkRkPVTzEqjJ2qKn4eUt)  \n- Reference Papers: [Vitta](https://arxiv.org/pdf/2408.05211), [Valle](https://github.com/lifeiteng/vall-e)  \n\n---\n\n### Русская Версия 🇷🇺\n\n---\n\n#### **Сравнение моделей**  \n**Метрики**:  \n- **PESQ@200**: Качество речи (чем выше, тем лучше).  \n- **STOI@200**: Разборчивость речи (ближе к 1 = лучше).  \n- **SI-SDR@200**: Соотношение сигнал-шум (выше = лучше).  \n- **SIM-O@200**: Сходство с эталоном (выше = лучше).  \n\n| Модель                   | PESQ@200       | STOI@200      | SI-SDR@200        | SIM-O@200      |  \n|--------------------------|----------------|---------------|-------------------|----------------|  \n| Original (LibriSpeech)   | 4.15           | 0.997         | 27.45 ±1.09       | —              |  \n| Parler TTS Mini v1       | 1.25 ±0.49     | 0.15 ±0.12    | 25.0 ±2.9         | 0.88 ±0.03     |  \n| Fish Speech 1.5          | 1.26 ±0.38     | 0.17 ±0.12    | 25.0 ±3.2         | 0.91 ±0.02     |  \n| **Salt-ASR Wav-Uni 1-12k **  | **1.27 ±0.40**     | 0.18 ±0.09    | 20.3 ±3.69        | 0.88 ±0.02     |  \n\n---\n\n#### **Наше решение**  \n- **Метод**: Расширение словаря LLM аудиотокенами + дообучение на **TTS** и **ASR**.  \n- **Обучение**:  \n  - SpeechTokenizer (семитические + аудиотокены) показал лучшие результаты, чем Encodec.  \n  - Время обучения: **150 часов на A100**.  \n- **Преимущества**: Единая функция потерь для двух задач, минимальные затраты.  \n\n\n---\n\n#### **Ресурсы**  \n- Код: [GitHub](https://github.com/VikhrModels/Vikhr4o)  \n- Демо: [Google Colab](https://colab.research.google.com/drive/1Poz6jNJu7-HRIkRkPVTzEqjJ2qKn4eUt)  \n\n--- \n\n**Примечание**: Модель поддерживает генерацию коротких фраз на английском, немецком и французском.\n\n",
    "related_quantizations": []
  },
  "tags": [
    "gguf",
    "arxiv:2408.05211",
    "endpoints_compatible",
    "region:us"
  ],
  "likes": 0,
  "downloads": 550,
  "gated": false,
  "private": false,
  "last_modified": "2025-03-28T07:52:17.000Z",
  "created_at": "2025-03-28T06:45:59.000Z",
  "pipeline_tag": "",
  "library_name": ""
}

{
  "_id": "67e645a7401e4ad0e91b1213",
  "id": "RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf",
  "modelId": "RichardErkhov/Vikhrmodels_-_salt-asr_wav-uni_1_tts_wav-uni_1-12k-gguf",
  "sha": "d3ee2aff61eae755e5d2412b1a6bb05043f2bbd7",
  "createdAt": "2025-03-28T06:45:59.000Z",
  "lastModified": "2025-03-28T07:52:17.000Z",
  "author": "RichardErkhov",
  "downloads": 550,
  "likes": 0,
  "gated": false,
  "private": false,
  "pipeline_tag": "",
  "library_name": "",
  "siblings_count": 24
}