GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
nineninesix
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
nineninesix
Model Tree
openai/whisper-small
Input Modalities
Output Modalities
Supported Functionality
kyrgyz-whisper-small is a fine-tuned multilingual speech recognition model based on OpenAI's Whisper Small architecture. This model adds native Kyrgyz language support while maintaining strong performance on English and Russian.
<|ky|> token.The following visualization shows improvement after fine-tuning:
Key Observations:
python
from transformers import AutoTokenizer # Load custom tokenizer with Kyrgyz supporttokenizer = AutoTokenizer.from_pretrained( "nineninesix/kyrgyz-whisper-small", trust_remote_code=True, ### !!! important !!! language="kyrgyz", task="transcribe")The <|ky|> token was initialized as an average of embeddings from linguistically similar languages:
python
embedding_ky = (embedding_ru + embedding_kk + embedding_tr) / 3python
import torchfrom transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline, WhisperFeatureExtractor, AutoTokenizer device = "cuda:0" if torch.cuda.is_available() else "cpu"torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32 model_id = "nineninesix/kyrgyz-whisper-small" model = AutoModelForSpeechSeq2Seq.from_pretrained( model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True)model.to(device) feature_extractor = WhisperFeatureExtractor.from_pretrained(model_id) tokenizer = AutoTokenizer.from_pretrained( model_id, trust_remote_code=True, language="kyrgyz", task="transcribe") pipe = pipeline( "automatic-speech-recognition", model=model, tokenizer=tokenizer, feature_extractor=feature_extractor, torch_dtype=torch_dtype, device=device) result = pipe("audio.mp3")print(result['text'])This model serves as a foundation for domain-specific fine-tuning using LoRA (Low-Rank Adaptation).
Unsloth integration example: see this Google Colab
Benefits of LoRA fine-tuning:
bibtex
@misc{kyrgyz-whisper-small, author = {nineninesix}, title = {Whisper Small - Kyrgyz, English, Russian}, year = {2025}, publisher = {HuggingFace}, url = {https://huggingface.co/nineninesix/kyrgyz-whisper-small}} @misc{radford2022whisper, doi = {10.48550/ARXIV.2212.04356}, url = {https://arxiv.org/abs/2212.04356}, author = {Radford, Alec and Kim, Jong Wook and Xu, Tao and Brockman, Greg and McLeavey, Christine and Sutskever, Ilya}, title = {Robust Speech Recognition via Large-Scale Weak Supervision}, publisher = {arXiv}, year = {2022}, copyright = {arXiv.org perpetual, non-exclusive license}}kyrgyz-ai/whisper_tokenizer_kyApache 2.0 - see LICENSE file for details.
from transformers import AutoTokenizer # Load custom tokenizer with Kyrgyz supporttokenizer = AutoTokenizer.from_pretrained( "nineninesix/kyrgyz-whisper-small", trust_remote_code=True, ### !!! important !!! language="kyrgyz", task="transcribe")import torchfrom transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline, WhisperFeatureExtractor, AutoTokenizer device = "cuda:0" if torch.cuda.is_available() else "cpu"torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32 model_id = "nineninesix/kyrgyz-whisper-small" model = AutoModelForSpeechSeq2Seq.from_pretrained( model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True)model.to(device) feature_extractor = WhisperFeatureExtractor.from_pretrained(model_id) tokenizer = AutoTokenizer.from_pretrained( model_id, trust_remote_code=True, language="kyrgyz", task="transcribe") pipe = pipeline( "automatic-speech-recognition", model=model, tokenizer=tokenizer, feature_extractor=feature_extractor, torch_dtype=torch_dtype, device=device) result = pipe("audio.mp3")print(result['text'])@misc{kyrgyz-whisper-small, author = {nineninesix}, title = {Whisper Small - Kyrgyz, English, Russian}, year = {2025}, publisher = {HuggingFace}, url = {https://huggingface.co/nineninesix/kyrgyz-whisper-small}} @misc{radford2022whisper, doi = {10.48550/ARXIV.2212.04356}, url = {https://arxiv.org/abs/2212.04356}, author = {Radford, Alec and Kim, Jong Wook and Xu, Tao and Brockman, Greg and McLeavey, Christine and Sutskever, Ilya}, title = {Robust Speech Recognition via Large-Scale Weak Supervision}, publisher = {arXiv}, year = {2022}, copyright = {arXiv.org perpetual, non-exclusive license}}