ModelPublicadoExterno

Ultravox Gemma 2 2b Jpn It

import transformers import librosa import torch import numpy as np from typing import Dict, Any

Fuentehuggingface.co/neody/ultravox-gemma-2-2b-jpn-it/tree/02cd5c3617f0b1a0b08d1064e9a87d1518bd1ea6

TipoOtherModelo basegoogle/gemma-2-2b-jpn-itVersiónRevision 02cd5c3Bcalidad

Fuente del modelo

Tipo: Other
Tamaño del archivo: 89,0 MB
Modelo base: google/gemma-2-2b-jpn-it
Versión: vRevision 02cd5c3
Actualizado: 31 oct 2026
Descargas: 53
Fuente: Hugging Face

Descripción de la fuente

import transformers
import librosa
import torch
import numpy as np
from typing import Dict, Any

model = transformers.AutoModel.from_pretrained(
    "neody/ultravox-gemma-2-2b-jpn-it", trust_remote_code=True
)
model.to("cuda", dtype=torch.bfloat16)
processor = transformers.AutoProcessor.from_pretrained(
    "neody/ultravox-gemma-2-2b-jpn-it", trust_remote_code=True
)
path = "record.wav"
audio, sr = librosa.load(path, sr=16000)


def preprocess(inputs: Dict[str, Any], device, dtype):
    turns: list = inputs.get("turns", [])

    audio = inputs.get("audio", None)
    # Convert to float32 if needed.
    if isinstance(audio, np.ndarray):
        if audio.dtype == np.float64:
            audio = audio.astype(np.float32)
        elif audio.dtype == np.int16:
            audio = audio.astype(np.float32) / np.float32(32768.0)
        elif audio.dtype == np.int32:
            audio = audio.astype(np.float32) / np.float32(2147483648.0)

    if audio is not None and (len(turns) == 0 or turns[-1]["role"] != "user"):
        prompt = inputs.get("prompt", "<|audio|>")
        if "<|audio|>" not in prompt:
            print(
                "Prompt does not contain '<|audio|>', appending '<|audio|>' to the end of the prompt."
            )

            prompt += " <|audio|>"
        turns.append({"role": "user", "content": prompt})

    text = processor.tokenizer.apply_chat_template(
        turns, add_generation_prompt=True, tokenize=False
    )

    if "sampling_rate" not in inputs and audio is not None:
        print(
            "No sampling rate provided, using default of 16kHz. We highly recommend providing the correct sampling rate."
        )

    output = processor(
        text=text,
        audio=audio,
        sampling_rate=inputs.get("sampling_rate", 16000),
    )
    if "audio_values" in output:
        output["audio_values"] = output["audio_values"].to(device, dtype)
    return output.to(device, dtype)


turns = []
print(
    processor.tokenizer.decode(
        model.generate(
            **preprocess(
                {"audio": audio, "turns": turns, "sampling_rate": sr},
                "cuda",
                torch.bfloat16,
            ),
            max_new_tokens=300,
        ).squeeze(),
        skip_special_tokens=True,
    )
)

Fuentes

1 fuente

Hugging FacePrincipalVersión 02cd5c3617f0b1a0b08d1064e9a87d1518bd1ea6

huggingface.co/neody/ultravox-gemma-2-2b-jpn-it/tree/02cd5c3617f0b1a0b08d1064e9a87d1518bd1ea6

Verificado 31 oct

Download files

Repository files

File inventory only; completeness is unverified.

Artefactos del modelo

1 artefacto

model.safetensors

safetensors · BF16 (unverified) · 89,0 MB · SHA-256 844a3876dbf4…187e · Hugging Face

Download file

Extractos de fuentes

3 extractos

Readme15 agohuggingface.co/neody/ultravox-gemma-2-2b-jpn-it

--- language: - ja base_model: - google/gemma-2-2b-jpn-it pipeline_tag: audio-text-to-text license: gemma datasets: - fixie-ai/common_voice_17_0 --- ```py import transformers import librosa import torch import numpy as np from typing import Dict, Any model = transformers.AutoModel.from_pretrained( "neody/ultravox-gemma-2-2b-jpn-it", trust_remote_code=True ) model.to("cuda", dtype=torch.bfloat16) processor = transformers.AutoProcessor.from_pretrained( "neody/ultravox-gemma-2-2b-jpn-it", trust_remote_code=True ) path = "record.wav" audio, sr = librosa.load(path, sr=16000) def preprocess(inputs: Dict[str, Any], device, dtype): turns: list = inputs.get("turns", []) audio = inputs.get("audio", None) # Convert to float32 if needed. if isinstance(audio, np.ndarray): if audio.dtype == np.float64: audio = audio.astype(np.float32) elif audio.dtype == np.int16: audio = audio.astype(np.float32) / np.float32(32768.0) elif audio.dtype == np.int32: audio = audio.astype(np.float32) / np.float32(2147483648.0) if audio is not None and (len(turns) == 0 or turns[-1]["role"] != "user"): prompt = inputs.get("prompt", "<|audio|>") if "<|audio|>" not in prompt: print( "Prompt does not contain '<|audio|>', appending '<|audio|>' to the end of the prompt." ) prompt += " <|audio|>" turns.append({"role": "user", "content": prompt}) text = processor.tokenizer.apply_chat_template( turns, add_generation_prompt=True, tokenize=False ) if "sampling_rate" not in inputs and audio is not None: print( "No sampling rate provided, using default of 16kHz. We highly recommend providing the correct sampling rate." ) output = processor( text=text, audio=audio, sampling_rate=inputs.get("sampling_rate", 16000), ) if "audio_values" in output: output["audio_values"] = output["audio_values"].to(device, dtype) return output.to(device, dtype) turns = [] print( processor.tokenizer.decode( model.generate( **preprocess( {"audio": audio, "turns": turns, "sampling_rate": sr}, "cuda", torch.bfloat16, ), max_new_tokens=300, ).squeeze(), skip_special_tokens=True, ) ) ```

Note31 octhuggingface.co/neody/ultravox-gemma-2-2b-jpn-it

Source context: 53 downloads · 1 likes · Pipeline audio-text-to-text · Repo neody/ultravox-gemma-2-2b-jpn-it

Note15 agohuggingface.co/neody/ultravox-gemma-2-2b-jpn-it

Source context: 26 downloads · 1 likes · Pipeline audio-text-to-text · Repo neody/ultravox-gemma-2-2b-jpn-it