Instructions to use aufklarer/Indic-Mio-MLX-fp16 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use aufklarer/Indic-Mio-MLX-fp16 with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download aufklarer/Indic-Mio-MLX-fp16 --local-dir Indic-Mio-MLX-fp16
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Upload Indic-Mio MLX fp16 bundle with MioCodec companion
Browse files- .gitattributes +1 -0
- README.md +131 -0
- added_tokens.json +0 -0
- bundle_config.json +26 -0
- chat_template.jinja +6 -0
- config.json +61 -0
- generation_config.json +11 -0
- merges.txt +0 -0
- miocodec/config.json +7 -0
- miocodec/config.yaml +98 -0
- miocodec/model.safetensors +3 -0
- miocodec/soniqo_manifest.json +61 -0
- model.safetensors +3 -0
- soniqo_manifest.json +148 -0
- special_tokens_map.json +31 -0
- tokenizer.json +3 -0
- tokenizer_config.json +0 -0
- vocab.json +0 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
tokenizer.json filter=lfs diff=lfs merge=lfs -text
|
README.md
ADDED
|
@@ -0,0 +1,131 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- hi
|
| 5 |
+
- bn
|
| 6 |
+
- mr
|
| 7 |
+
- te
|
| 8 |
+
- kn
|
| 9 |
+
- mai
|
| 10 |
+
- as
|
| 11 |
+
- gu
|
| 12 |
+
- ml
|
| 13 |
+
- pa
|
| 14 |
+
- ta
|
| 15 |
+
- ne
|
| 16 |
+
- sa
|
| 17 |
+
- en
|
| 18 |
+
tags:
|
| 19 |
+
- text-to-speech
|
| 20 |
+
- hindi
|
| 21 |
+
- indic
|
| 22 |
+
- emotion-control
|
| 23 |
+
- voice-cloning
|
| 24 |
+
- mlx
|
| 25 |
+
- fp16
|
| 26 |
+
base_model:
|
| 27 |
+
- SPRINGLab/Indic-Mio
|
| 28 |
+
- Aratako/MioCodec-25Hz-24kHz
|
| 29 |
+
library_name: mlx
|
| 30 |
+
pipeline_tag: text-to-speech
|
| 31 |
+
---
|
| 32 |
+
|
| 33 |
+
# Indic-Mio-MLX-fp16
|
| 34 |
+
|
| 35 |
+
Apple Silicon fp16 export of [SPRINGLab/Indic-Mio](https://huggingface.co/SPRINGLab/Indic-Mio), bundled with the [MioCodec 25 Hz / 24 kHz](https://huggingface.co/Aratako/MioCodec-25Hz-24kHz) companion required to decode Indic-Mio content tokens into waveform audio.
|
| 36 |
+
|
| 37 |
+
This is an artifact bundle for Swift/MLX runtime work. It preserves upstream key names; runtime code is expected to implement model loading, sampling, MioCodec conditioning, and waveform decoding.
|
| 38 |
+
|
| 39 |
+
## Model
|
| 40 |
+
|
| 41 |
+
| Field | Value |
|
| 42 |
+
|---|---|
|
| 43 |
+
| Primary source | `SPRINGLab/Indic-Mio` |
|
| 44 |
+
| Primary source revision | `25feace00ca76c71c40b1e8d921fd3c2943c545e` |
|
| 45 |
+
| Bundled codec source | `Aratako/MioCodec-25Hz-24kHz` |
|
| 46 |
+
| Bundled codec revision | `3a737f0de2c6324cb2fe40c1fbd1056c7add423d` |
|
| 47 |
+
| Format | MLX fp16 safetensors |
|
| 48 |
+
| Primary license | Apache-2.0 |
|
| 49 |
+
| Codec license | MIT |
|
| 50 |
+
| Total bundle size | 1.5 GB |
|
| 51 |
+
| Sample rate | 24,000 Hz |
|
| 52 |
+
| Voice conditioning | MioCodec global embeddings / zero-shot voice path |
|
| 53 |
+
| Runtime status | Artifact export; Swift/MLX runtime port required |
|
| 54 |
+
|
| 55 |
+
## Emotion Control
|
| 56 |
+
|
| 57 |
+
| Field | Value |
|
| 58 |
+
|---|---|
|
| 59 |
+
| Marker syntax | Suffix angle tags |
|
| 60 |
+
| Supported markers | `<happy>`, `<sad>`, `<angry>`, `<disgust>`, `<fear>`, `<surprise>` |
|
| 61 |
+
|
| 62 |
+
## Files
|
| 63 |
+
|
| 64 |
+
| File | Size | Description |
|
| 65 |
+
|---|---:|---|
|
| 66 |
+
| `README.md` | 2.3 KB | Model card. |
|
| 67 |
+
| `added_tokens.json` | 309.6 KB | Tokenizer asset copied from the upstream Indic-Mio repository. |
|
| 68 |
+
| `bundle_config.json` | 697 B | Soniqo bundle metadata linking the text model and bundled codec. |
|
| 69 |
+
| `chat_template.jinja` | 205 B | Upstream generation/template metadata. |
|
| 70 |
+
| `config.json` | 1.4 KB | Upstream Qwen/Indic-Mio model config at repo root for runtime loading and HF download tracking. |
|
| 71 |
+
| `generation_config.json` | 183 B | Upstream generation/template metadata. |
|
| 72 |
+
| `merges.txt` | 1.7 MB | Tokenizer asset copied from the upstream Indic-Mio repository. |
|
| 73 |
+
| `miocodec/config.json` | 230 B | MioCodec runtime/export config. |
|
| 74 |
+
| `miocodec/config.yaml` | 2.4 KB | MioCodec runtime/export config. |
|
| 75 |
+
| `miocodec/model.safetensors` | 261.6 MB | Bundled MioCodec fp16 weights required for waveform decoding. |
|
| 76 |
+
| `miocodec/soniqo_manifest.json` | 1.3 KB | Export manifest with source revisions, license posture, and bundle metadata. |
|
| 77 |
+
| `model.safetensors` | 1.2 GB | Indic-Mio fp16 weights with upstream key names preserved. |
|
| 78 |
+
| `soniqo_manifest.json` | 3.1 KB | Export manifest with source revisions, license posture, and bundle metadata. |
|
| 79 |
+
| `special_tokens_map.json` | 613 B | Tokenizer asset copied from the upstream Indic-Mio repository. |
|
| 80 |
+
| `tokenizer.json` | 13.8 MB | Tokenizer asset copied from the upstream Indic-Mio repository. |
|
| 81 |
+
| `tokenizer_config.json` | 2.3 MB | Tokenizer asset copied from the upstream Indic-Mio repository. |
|
| 82 |
+
| `vocab.json` | 2.8 MB | Tokenizer asset copied from the upstream Indic-Mio repository. |
|
| 83 |
+
|
| 84 |
+
## Validation
|
| 85 |
+
|
| 86 |
+
| Test | Result |
|
| 87 |
+
|---|---|
|
| 88 |
+
| Indic-Mio + MioCodec local roundtrip | Passed |
|
| 89 |
+
| Input text | `नमस्ते, मैं आज बहुत खुश हूँ। <happy>` |
|
| 90 |
+
| ASR transcript | `नमस्ते, मैं आज बहुत खुश हूँ` |
|
| 91 |
+
| Sample rate | 24000 Hz |
|
| 92 |
+
|
| 93 |
+
The roundtrip test was a local synthesis and transcription sanity check. It verifies that the exported Indic-Mio weights and bundled MioCodec path can produce intelligible Hindi speech for an explicit emotion marker. It is not a full runtime benchmark.
|
| 94 |
+
|
| 95 |
+
## Usage
|
| 96 |
+
|
| 97 |
+
These artifacts are not a drop-in Transformers checkpoint. They are intended for `speech-swift` or another MLX runtime that implements the Indic-Mio token generation path and MioCodec decoding path.
|
| 98 |
+
|
| 99 |
+
```bash
|
| 100 |
+
git lfs install
|
| 101 |
+
git clone https://huggingface.co/aufklarer/Indic-Mio-MLX-fp16
|
| 102 |
+
```
|
| 103 |
+
|
| 104 |
+
Expected bundle layout:
|
| 105 |
+
|
| 106 |
+
```text
|
| 107 |
+
Indic-Mio-MLX-fp16/
|
| 108 |
+
model.safetensors
|
| 109 |
+
config.json
|
| 110 |
+
tokenizer.json
|
| 111 |
+
bundle_config.json
|
| 112 |
+
soniqo_manifest.json
|
| 113 |
+
miocodec/
|
| 114 |
+
model.safetensors
|
| 115 |
+
config.json
|
| 116 |
+
config.yaml
|
| 117 |
+
soniqo_manifest.json
|
| 118 |
+
```
|
| 119 |
+
|
| 120 |
+
## Source
|
| 121 |
+
|
| 122 |
+
- Indic-Mio upstream model: https://huggingface.co/SPRINGLab/Indic-Mio
|
| 123 |
+
- MioCodec upstream model: https://huggingface.co/Aratako/MioCodec-25Hz-24kHz
|
| 124 |
+
|
| 125 |
+
## Links
|
| 126 |
+
|
| 127 |
+
- [speech-swift](https://github.com/soniqo/speech-swift) - Apple SDK
|
| 128 |
+
- [Speech Studio](https://soniqo.audio/speech-studio) - local speech generation and voice cloning app
|
| 129 |
+
- [Docs](https://soniqo.audio/getting-started) - install and CLI docs
|
| 130 |
+
- [soniqo.audio](https://soniqo.audio) - website
|
| 131 |
+
- [Blog](https://soniqo.audio/blog) - blog
|
added_tokens.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
bundle_config.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"bundle_type": "hindi_emotion_tts_with_codec",
|
| 4 |
+
"primary_model": {
|
| 5 |
+
"path": ".",
|
| 6 |
+
"source_repo": "SPRINGLab/Indic-Mio",
|
| 7 |
+
"source_revision": "25feace00ca76c71c40b1e8d921fd3c2943c545e",
|
| 8 |
+
"format": "MLX fp16 safetensors",
|
| 9 |
+
"emotion_markers": [
|
| 10 |
+
"<happy>",
|
| 11 |
+
"<sad>",
|
| 12 |
+
"<angry>",
|
| 13 |
+
"<disgust>",
|
| 14 |
+
"<fear>",
|
| 15 |
+
"<surprise>"
|
| 16 |
+
]
|
| 17 |
+
},
|
| 18 |
+
"codec": {
|
| 19 |
+
"path": "miocodec/",
|
| 20 |
+
"source_repo": "Aratako/MioCodec-25Hz-24kHz",
|
| 21 |
+
"source_revision": "3a737f0de2c6324cb2fe40c1fbd1056c7add423d",
|
| 22 |
+
"format": "MLX fp16 safetensors",
|
| 23 |
+
"sample_rate_hz": 24000
|
| 24 |
+
},
|
| 25 |
+
"runtime_status": "artifact-export; Swift/MLX runtime port required"
|
| 26 |
+
}
|
chat_template.jinja
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{%- for message in messages %}
|
| 2 |
+
{{ '<|im_start|>' + message['role'] + '\n' + message['content'] + '<|im_end|>\n' }}
|
| 3 |
+
{%- endfor %}
|
| 4 |
+
{%- if add_generation_prompt %}
|
| 5 |
+
{{ '<|im_start|>assistant\n' }}
|
| 6 |
+
{%- endif %}
|
config.json
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"Qwen3ForCausalLM"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"dtype": "bfloat16",
|
| 8 |
+
"eos_token_id": 151645,
|
| 9 |
+
"head_dim": 128,
|
| 10 |
+
"hidden_act": "silu",
|
| 11 |
+
"hidden_size": 1024,
|
| 12 |
+
"initializer_range": 0.02,
|
| 13 |
+
"intermediate_size": 3072,
|
| 14 |
+
"layer_types": [
|
| 15 |
+
"full_attention",
|
| 16 |
+
"full_attention",
|
| 17 |
+
"full_attention",
|
| 18 |
+
"full_attention",
|
| 19 |
+
"full_attention",
|
| 20 |
+
"full_attention",
|
| 21 |
+
"full_attention",
|
| 22 |
+
"full_attention",
|
| 23 |
+
"full_attention",
|
| 24 |
+
"full_attention",
|
| 25 |
+
"full_attention",
|
| 26 |
+
"full_attention",
|
| 27 |
+
"full_attention",
|
| 28 |
+
"full_attention",
|
| 29 |
+
"full_attention",
|
| 30 |
+
"full_attention",
|
| 31 |
+
"full_attention",
|
| 32 |
+
"full_attention",
|
| 33 |
+
"full_attention",
|
| 34 |
+
"full_attention",
|
| 35 |
+
"full_attention",
|
| 36 |
+
"full_attention",
|
| 37 |
+
"full_attention",
|
| 38 |
+
"full_attention",
|
| 39 |
+
"full_attention",
|
| 40 |
+
"full_attention",
|
| 41 |
+
"full_attention",
|
| 42 |
+
"full_attention"
|
| 43 |
+
],
|
| 44 |
+
"max_position_embeddings": 32768,
|
| 45 |
+
"max_window_layers": 28,
|
| 46 |
+
"model_type": "qwen3",
|
| 47 |
+
"num_attention_heads": 16,
|
| 48 |
+
"num_hidden_layers": 28,
|
| 49 |
+
"num_key_value_heads": 8,
|
| 50 |
+
"pad_token_id": 151643,
|
| 51 |
+
"rms_norm_eps": 1e-06,
|
| 52 |
+
"rope_scaling": null,
|
| 53 |
+
"rope_theta": 1000000,
|
| 54 |
+
"sliding_window": null,
|
| 55 |
+
"tie_word_embeddings": true,
|
| 56 |
+
"transformers_version": "4.57.6",
|
| 57 |
+
"unsloth_version": "2026.2.1",
|
| 58 |
+
"use_cache": false,
|
| 59 |
+
"use_sliding_window": false,
|
| 60 |
+
"vocab_size": 164480
|
| 61 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_sample": true,
|
| 3 |
+
"eos_token_id": [
|
| 4 |
+
151645,
|
| 5 |
+
151643
|
| 6 |
+
],
|
| 7 |
+
"max_length": 32768,
|
| 8 |
+
"max_new_tokens": 2048,
|
| 9 |
+
"pad_token_id": 151643,
|
| 10 |
+
"transformers_version": "4.57.6"
|
| 11 |
+
}
|
merges.txt
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
miocodec/config.json
ADDED
|
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "mio-codec-25hz-24khz",
|
| 3 |
+
"source_repo": "Aratako/MioCodec-25Hz-24kHz",
|
| 4 |
+
"format": "MLX fp16 safetensors",
|
| 5 |
+
"sample_rate_hz": 24000,
|
| 6 |
+
"runtime_status": "artifact-export; Swift codec runtime not implemented"
|
| 7 |
+
}
|
miocodec/config.yaml
ADDED
|
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
model:
|
| 2 |
+
class_path: miocodec.model.MioCodecModel
|
| 3 |
+
init_args:
|
| 4 |
+
config:
|
| 5 |
+
# SSL Feature settings
|
| 6 |
+
local_ssl_layers: [6, 9]
|
| 7 |
+
global_ssl_layers: [1, 2]
|
| 8 |
+
normalize_ssl_features: true
|
| 9 |
+
|
| 10 |
+
# Down/up-sampling settings
|
| 11 |
+
downsample_factor: 2
|
| 12 |
+
use_conv_downsample: true
|
| 13 |
+
|
| 14 |
+
# Audio settings
|
| 15 |
+
sample_rate: 24000
|
| 16 |
+
n_fft: 1920
|
| 17 |
+
hop_length: 480
|
| 18 |
+
|
| 19 |
+
# Wave decoder settings
|
| 20 |
+
use_wave_decoder: true
|
| 21 |
+
wave_upsample_factor: 2
|
| 22 |
+
wave_interpolation_mode: linear
|
| 23 |
+
wave_decoder_dim: 512
|
| 24 |
+
wave_resnet_num_blocks: 2
|
| 25 |
+
wave_resnet_kernel_size: 3
|
| 26 |
+
wave_resnet_num_groups: 32
|
| 27 |
+
wave_resnet_dropout: 0.1
|
| 28 |
+
istft_padding: same
|
| 29 |
+
|
| 30 |
+
ssl_feature_extractor:
|
| 31 |
+
class_path: miocodec.module.ssl_extractor.SSLFeatureExtractor
|
| 32 |
+
init_args:
|
| 33 |
+
model_name: wavlm_base_plus
|
| 34 |
+
output_layer: 9
|
| 35 |
+
sample_rate: 24000
|
| 36 |
+
|
| 37 |
+
local_encoder:
|
| 38 |
+
class_path: miocodec.module.transformer.Transformer
|
| 39 |
+
init_args:
|
| 40 |
+
dim: 768
|
| 41 |
+
n_layers: 6
|
| 42 |
+
n_heads: 12
|
| 43 |
+
window_size: 125
|
| 44 |
+
use_rope: true
|
| 45 |
+
rope_theta: 10000.0
|
| 46 |
+
max_seq_len: 512
|
| 47 |
+
use_flash_attention: true
|
| 48 |
+
|
| 49 |
+
local_quantizer:
|
| 50 |
+
class_path: miocodec.module.fsq.FiniteScalarQuantizer
|
| 51 |
+
init_args:
|
| 52 |
+
input_dim: 768
|
| 53 |
+
output_dim: 768
|
| 54 |
+
levels: [8, 8, 8, 5, 5] # 12800
|
| 55 |
+
|
| 56 |
+
feature_decoder: null
|
| 57 |
+
|
| 58 |
+
global_encoder:
|
| 59 |
+
class_path: miocodec.module.global_encoder.GlobalEncoder
|
| 60 |
+
init_args:
|
| 61 |
+
input_channels: 768
|
| 62 |
+
output_channels: 128
|
| 63 |
+
num_layers: 4
|
| 64 |
+
dim: 384
|
| 65 |
+
intermediate_dim: 1152
|
| 66 |
+
|
| 67 |
+
# Mel decoder not used
|
| 68 |
+
mel_prenet: null
|
| 69 |
+
mel_decoder: null
|
| 70 |
+
mel_postnet: null
|
| 71 |
+
|
| 72 |
+
# Wave decoder components
|
| 73 |
+
wave_prenet:
|
| 74 |
+
class_path: miocodec.module.transformer.Transformer
|
| 75 |
+
init_args:
|
| 76 |
+
dim: 768
|
| 77 |
+
output_dim: 512
|
| 78 |
+
n_layers: 6
|
| 79 |
+
n_heads: 12
|
| 80 |
+
window_size: 65
|
| 81 |
+
use_rope: true
|
| 82 |
+
rope_theta: 10000.0
|
| 83 |
+
max_seq_len: 512
|
| 84 |
+
use_flash_attention: true
|
| 85 |
+
|
| 86 |
+
wave_decoder:
|
| 87 |
+
class_path: miocodec.module.transformer.Transformer
|
| 88 |
+
init_args:
|
| 89 |
+
dim: 512
|
| 90 |
+
n_layers: 8
|
| 91 |
+
n_heads: 8
|
| 92 |
+
window_size: 65
|
| 93 |
+
use_rope: true
|
| 94 |
+
rope_theta: 10000.0
|
| 95 |
+
max_seq_len: 512
|
| 96 |
+
adanorm_condition_dim: 128
|
| 97 |
+
use_adaln_zero: true
|
| 98 |
+
use_flash_attention: true
|
miocodec/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:97311822accd32928ebca51288aa03ed2a77c83efda953243dc3e298c6a8a064
|
| 3 |
+
size 261562102
|
miocodec/soniqo_manifest.json
ADDED
|
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"artifact_type": "hindi_emotion_tts_candidate",
|
| 4 |
+
"source_repo": "Aratako/MioCodec-25Hz-24kHz",
|
| 5 |
+
"source_revision": "3a737f0de2c6324cb2fe40c1fbd1056c7add423d",
|
| 6 |
+
"output_name": "MioCodec-25Hz-24kHz-MLX-fp16",
|
| 7 |
+
"license": "mit",
|
| 8 |
+
"license_posture": "commercial-safe",
|
| 9 |
+
"readiness": "required-companion",
|
| 10 |
+
"format": "MLX fp16 safetensors",
|
| 11 |
+
"runtime_status": "artifact-export; Swift codec runtime not implemented",
|
| 12 |
+
"language": [
|
| 13 |
+
"en",
|
| 14 |
+
"ja",
|
| 15 |
+
"nl",
|
| 16 |
+
"fr",
|
| 17 |
+
"de",
|
| 18 |
+
"it",
|
| 19 |
+
"pl",
|
| 20 |
+
"pt",
|
| 21 |
+
"es",
|
| 22 |
+
"ko",
|
| 23 |
+
"zh"
|
| 24 |
+
],
|
| 25 |
+
"sample_rate_hz": 24000,
|
| 26 |
+
"emotion_control": {
|
| 27 |
+
"marker_syntax": "none",
|
| 28 |
+
"markers": []
|
| 29 |
+
},
|
| 30 |
+
"voice_conditioning": "content tokens plus global embedding",
|
| 31 |
+
"required_companions": [],
|
| 32 |
+
"notes": [
|
| 33 |
+
"Required companion for Indic-Mio waveform decoding.",
|
| 34 |
+
"Upstream codec card describes content tokens plus global embeddings for zero-shot voice conversion."
|
| 35 |
+
],
|
| 36 |
+
"converted_files": [
|
| 37 |
+
"model.safetensors"
|
| 38 |
+
],
|
| 39 |
+
"copied_files": [
|
| 40 |
+
"README.md",
|
| 41 |
+
"config.yaml"
|
| 42 |
+
],
|
| 43 |
+
"files": [
|
| 44 |
+
{
|
| 45 |
+
"path": "README.md",
|
| 46 |
+
"bytes": 7545
|
| 47 |
+
},
|
| 48 |
+
{
|
| 49 |
+
"path": "config.json",
|
| 50 |
+
"bytes": 230
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"path": "config.yaml",
|
| 54 |
+
"bytes": 2404
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"path": "model.safetensors",
|
| 58 |
+
"bytes": 261562102
|
| 59 |
+
}
|
| 60 |
+
]
|
| 61 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b1bea5934460ccd06b36ff62b9f447792157e9b3f815295bd734193ea2520c71
|
| 3 |
+
size 1217824912
|
soniqo_manifest.json
ADDED
|
@@ -0,0 +1,148 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 1,
|
| 3 |
+
"artifact_type": "hindi_emotion_tts_candidate",
|
| 4 |
+
"source_repo": "SPRINGLab/Indic-Mio",
|
| 5 |
+
"source_revision": "25feace00ca76c71c40b1e8d921fd3c2943c545e",
|
| 6 |
+
"output_name": "Indic-Mio-MLX-fp16",
|
| 7 |
+
"license": "apache-2.0",
|
| 8 |
+
"license_posture": "commercial-safe",
|
| 9 |
+
"readiness": "primary-port-candidate",
|
| 10 |
+
"format": "MLX fp16 safetensors",
|
| 11 |
+
"runtime_status": "artifact-export; Swift runtime not implemented",
|
| 12 |
+
"language": [
|
| 13 |
+
"hi",
|
| 14 |
+
"bn",
|
| 15 |
+
"mr",
|
| 16 |
+
"te",
|
| 17 |
+
"kn",
|
| 18 |
+
"mai",
|
| 19 |
+
"as",
|
| 20 |
+
"gu",
|
| 21 |
+
"ml",
|
| 22 |
+
"pa",
|
| 23 |
+
"ta",
|
| 24 |
+
"ne",
|
| 25 |
+
"sa",
|
| 26 |
+
"en"
|
| 27 |
+
],
|
| 28 |
+
"sample_rate_hz": 24000,
|
| 29 |
+
"emotion_control": {
|
| 30 |
+
"marker_syntax": "suffix angle tags",
|
| 31 |
+
"markers": [
|
| 32 |
+
"<happy>",
|
| 33 |
+
"<sad>",
|
| 34 |
+
"<angry>",
|
| 35 |
+
"<disgust>",
|
| 36 |
+
"<fear>",
|
| 37 |
+
"<surprise>"
|
| 38 |
+
]
|
| 39 |
+
},
|
| 40 |
+
"voice_conditioning": "MioCodec global embeddings / zero-shot voice path",
|
| 41 |
+
"required_companions": [
|
| 42 |
+
"mio-codec-25hz-24khz"
|
| 43 |
+
],
|
| 44 |
+
"notes": [
|
| 45 |
+
"Primary Hindi emotion candidate.",
|
| 46 |
+
"Generates MioCodec content tokens; bundle needs the MioCodec companion for waveform decoding.",
|
| 47 |
+
"Upstream card documents end-of-utterance tags for Indian languages."
|
| 48 |
+
],
|
| 49 |
+
"converted_files": [
|
| 50 |
+
"model.safetensors"
|
| 51 |
+
],
|
| 52 |
+
"copied_files": [
|
| 53 |
+
"README.md",
|
| 54 |
+
"added_tokens.json",
|
| 55 |
+
"chat_template.jinja",
|
| 56 |
+
"config.json",
|
| 57 |
+
"generation_config.json",
|
| 58 |
+
"merges.txt",
|
| 59 |
+
"special_tokens_map.json",
|
| 60 |
+
"tokenizer.json",
|
| 61 |
+
"tokenizer_config.json",
|
| 62 |
+
"vocab.json"
|
| 63 |
+
],
|
| 64 |
+
"files": [
|
| 65 |
+
{
|
| 66 |
+
"path": "README.md",
|
| 67 |
+
"bytes": 2268
|
| 68 |
+
},
|
| 69 |
+
{
|
| 70 |
+
"path": "added_tokens.json",
|
| 71 |
+
"bytes": 309597
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"path": "bundle_config.json",
|
| 75 |
+
"bytes": 697
|
| 76 |
+
},
|
| 77 |
+
{
|
| 78 |
+
"path": "chat_template.jinja",
|
| 79 |
+
"bytes": 205
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"path": "config.json",
|
| 83 |
+
"bytes": 1394
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"path": "generation_config.json",
|
| 87 |
+
"bytes": 183
|
| 88 |
+
},
|
| 89 |
+
{
|
| 90 |
+
"path": "merges.txt",
|
| 91 |
+
"bytes": 1671853
|
| 92 |
+
},
|
| 93 |
+
{
|
| 94 |
+
"path": "miocodec/config.json",
|
| 95 |
+
"bytes": 230
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"path": "miocodec/config.yaml",
|
| 99 |
+
"bytes": 2404
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"path": "miocodec/model.safetensors",
|
| 103 |
+
"bytes": 261562102
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"path": "miocodec/soniqo_manifest.json",
|
| 107 |
+
"bytes": 1340
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"path": "model.safetensors",
|
| 111 |
+
"bytes": 1217824912
|
| 112 |
+
},
|
| 113 |
+
{
|
| 114 |
+
"path": "soniqo_manifest.json",
|
| 115 |
+
"bytes": 3079
|
| 116 |
+
},
|
| 117 |
+
{
|
| 118 |
+
"path": "special_tokens_map.json",
|
| 119 |
+
"bytes": 613
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"path": "tokenizer.json",
|
| 123 |
+
"bytes": 13817944
|
| 124 |
+
},
|
| 125 |
+
{
|
| 126 |
+
"path": "tokenizer_config.json",
|
| 127 |
+
"bytes": 2272720
|
| 128 |
+
},
|
| 129 |
+
{
|
| 130 |
+
"path": "vocab.json",
|
| 131 |
+
"bytes": 2776833
|
| 132 |
+
}
|
| 133 |
+
],
|
| 134 |
+
"bundle_layout": {
|
| 135 |
+
"primary_model": ".",
|
| 136 |
+
"bundled_companions": [
|
| 137 |
+
{
|
| 138 |
+
"name": "MioCodec-25Hz-24kHz-MLX-fp16",
|
| 139 |
+
"path": "miocodec/",
|
| 140 |
+
"source_repo": "Aratako/MioCodec-25Hz-24kHz",
|
| 141 |
+
"source_revision": "3a737f0de2c6324cb2fe40c1fbd1056c7add423d",
|
| 142 |
+
"license": "mit",
|
| 143 |
+
"sample_rate_hz": 24000,
|
| 144 |
+
"purpose": "Decode Indic-Mio content tokens into waveform audio."
|
| 145 |
+
}
|
| 146 |
+
]
|
| 147 |
+
}
|
| 148 |
+
}
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"additional_special_tokens": [
|
| 3 |
+
"<|im_start|>",
|
| 4 |
+
"<|im_end|>",
|
| 5 |
+
"<|object_ref_start|>",
|
| 6 |
+
"<|object_ref_end|>",
|
| 7 |
+
"<|box_start|>",
|
| 8 |
+
"<|box_end|>",
|
| 9 |
+
"<|quad_start|>",
|
| 10 |
+
"<|quad_end|>",
|
| 11 |
+
"<|vision_start|>",
|
| 12 |
+
"<|vision_end|>",
|
| 13 |
+
"<|vision_pad|>",
|
| 14 |
+
"<|image_pad|>",
|
| 15 |
+
"<|video_pad|>"
|
| 16 |
+
],
|
| 17 |
+
"eos_token": {
|
| 18 |
+
"content": "<|im_end|>",
|
| 19 |
+
"lstrip": false,
|
| 20 |
+
"normalized": false,
|
| 21 |
+
"rstrip": false,
|
| 22 |
+
"single_word": false
|
| 23 |
+
},
|
| 24 |
+
"pad_token": {
|
| 25 |
+
"content": "<|endoftext|>",
|
| 26 |
+
"lstrip": false,
|
| 27 |
+
"normalized": false,
|
| 28 |
+
"rstrip": false,
|
| 29 |
+
"single_word": false
|
| 30 |
+
}
|
| 31 |
+
}
|
tokenizer.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:abcde038b87ccd029a4523b0c5cec1da6d84b4f3d68b351495df086d63033f1f
|
| 3 |
+
size 13817944
|
tokenizer_config.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
vocab.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|