Instructions to use openlimi/Tunar-stt-tiny with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use openlimi/Tunar-stt-tiny with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="openlimi/Tunar-stt-tiny")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForSpeechSeq2Seq processor = AutoProcessor.from_pretrained("openlimi/Tunar-stt-tiny") model = AutoModelForSpeechSeq2Seq.from_pretrained("openlimi/Tunar-stt-tiny", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Upload Tunar-stt-tiny model from Google Drive
Browse files- README.md +35 -0
- config.json +36 -0
- generation_config.json +9 -0
- model.safetensors +3 -0
- preprocessor_config.json +9 -0
- tokenizer.json +0 -0
- tokenizer_config.json +9 -0
README.md
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- uz
|
| 4 |
+
library_name: transformers
|
| 5 |
+
base_model: moonshine-ai/moonshine-tiny
|
| 6 |
+
tags:
|
| 7 |
+
- automatic-speech-recognition
|
| 8 |
+
- moonshine
|
| 9 |
+
- uzbek
|
| 10 |
+
pipeline_tag: automatic-speech-recognition
|
| 11 |
+
---
|
| 12 |
+
# Tunar-stt-tiny
|
| 13 |
+
|
| 14 |
+
Uzbek speech recognition model fine-tuned from Moonshine-tiny (v3_1500h run).
|
| 15 |
+
|
| 16 |
+
- Output alphabet: **new Uzbek Latin alphabet** (`ö ğ ş ç`); convert with `o'->ö, g'->ğ, sh->ş, ch->ç`.
|
| 17 |
+
- Custom 4096-token BPE tokenizer.
|
| 18 |
+
- Input: 16 kHz mono audio, up to ~30 s per segment.
|
| 19 |
+
|
| 20 |
+
## Usage
|
| 21 |
+
```python
|
| 22 |
+
import torch, soundfile as sf
|
| 23 |
+
from transformers import MoonshineForConditionalGeneration, AutoTokenizer, AutoFeatureExtractor
|
| 24 |
+
|
| 25 |
+
repo = "openlimi/Tunar-stt-tiny"
|
| 26 |
+
model = MoonshineForConditionalGeneration.from_pretrained(repo).eval()
|
| 27 |
+
tok = AutoTokenizer.from_pretrained(repo)
|
| 28 |
+
fe = AutoFeatureExtractor.from_pretrained(repo)
|
| 29 |
+
|
| 30 |
+
wav, sr = sf.read("audio.wav") # 16 kHz mono
|
| 31 |
+
inputs = fe(wav, sampling_rate=16000, return_tensors="pt", return_attention_mask=True)
|
| 32 |
+
with torch.no_grad():
|
| 33 |
+
ids = model.generate(**inputs, max_new_tokens=int(len(wav) / 16000 * 12) + 8)
|
| 34 |
+
print(tok.batch_decode(ids, skip_special_tokens=True)[0])
|
| 35 |
+
```
|
config.json
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"MoonshineForConditionalGeneration"
|
| 4 |
+
],
|
| 5 |
+
"attention_bias": false,
|
| 6 |
+
"attention_dropout": 0.0,
|
| 7 |
+
"bos_token_id": 1,
|
| 8 |
+
"decoder_hidden_act": "silu",
|
| 9 |
+
"decoder_num_attention_heads": 8,
|
| 10 |
+
"decoder_num_hidden_layers": 6,
|
| 11 |
+
"decoder_num_key_value_heads": 8,
|
| 12 |
+
"decoder_start_token_id": 1,
|
| 13 |
+
"dtype": "float32",
|
| 14 |
+
"encoder_hidden_act": "gelu",
|
| 15 |
+
"encoder_num_attention_heads": 8,
|
| 16 |
+
"encoder_num_hidden_layers": 6,
|
| 17 |
+
"encoder_num_key_value_heads": 8,
|
| 18 |
+
"eos_token_id": 2,
|
| 19 |
+
"hidden_size": 288,
|
| 20 |
+
"initializer_range": 0.02,
|
| 21 |
+
"intermediate_size": 1152,
|
| 22 |
+
"is_encoder_decoder": true,
|
| 23 |
+
"max_position_embeddings": 194,
|
| 24 |
+
"model_type": "moonshine",
|
| 25 |
+
"pad_head_dim_to_multiple_of": 8,
|
| 26 |
+
"pad_token_id": 3,
|
| 27 |
+
"rope_parameters": {
|
| 28 |
+
"partial_rotary_factor": 0.9,
|
| 29 |
+
"rope_theta": 10000.0,
|
| 30 |
+
"rope_type": "default"
|
| 31 |
+
},
|
| 32 |
+
"tie_word_embeddings": true,
|
| 33 |
+
"transformers_version": "5.23.0",
|
| 34 |
+
"use_cache": true,
|
| 35 |
+
"vocab_size": 4096
|
| 36 |
+
}
|
generation_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_from_model_config": true,
|
| 3 |
+
"bos_token_id": 1,
|
| 4 |
+
"decoder_start_token_id": 1,
|
| 5 |
+
"eos_token_id": 2,
|
| 6 |
+
"max_length": 194,
|
| 7 |
+
"pad_token_id": 3,
|
| 8 |
+
"transformers_version": "5.23.0"
|
| 9 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb65079a0d9089513da766ab0734305a8eb0dbed5470fd75a810e7ffb73f3207
|
| 3 |
+
size 75358984
|
preprocessor_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"do_normalize": false,
|
| 3 |
+
"feature_extractor_type": "Wav2Vec2FeatureExtractor",
|
| 4 |
+
"feature_size": 1,
|
| 5 |
+
"padding_side": "right",
|
| 6 |
+
"padding_value": 0.0,
|
| 7 |
+
"return_attention_mask": true,
|
| 8 |
+
"sampling_rate": 16000
|
| 9 |
+
}
|
tokenizer.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backend": "tokenizers",
|
| 3 |
+
"bos_token": "<s>",
|
| 4 |
+
"eos_token": "</s>",
|
| 5 |
+
"model_max_length": 1000000000000000019884624838656,
|
| 6 |
+
"pad_token": "<pad>",
|
| 7 |
+
"tokenizer_class": "TokenizersBackend",
|
| 8 |
+
"unk_token": "<unk>"
|
| 9 |
+
}
|