__init__.py

"""
CUDA: 12.1
cuDNN Version: 8.9.2.26_1.0-1_amd64
Tensorflow Version: 2.12.0
Torch Version: 2.1.0.dev20230606+cu121
Transformers Version: 4.30.2
BENCHMARK:
    - RAM: 2.8 GB
    - VRAM: 1812 MB
    - test.wav: 23 s 
        - GPU (3060) -> 1.1s    (TensorCore is used for fp16 inference)
        - GPU (1660S) -> 3.3s
        - CPU -> torch.float16 not supported on CPU (Ryzen 5 3600)
    - Punchuation: True
"""

from transformers import (
    WhisperForConditionalGeneration, WhisperProcessor, WhisperConfig
)
import torch
import ffmpeg
import torch
import torch.nn.functional as F
import numpy as np
import os

# load_audio and pad_or_trim functions
SAMPLE_RATE = 16000
CHUNK_LENGTH = 30  # 30-second chunks
N_SAMPLES = CHUNK_LENGTH * SAMPLE_RATE  # 480000 samples in a 30-second chunk

class Model:
    def __init__(self, 
                 model_name_or_path: str, 
                 cuda_visible_device: str = "0", 
                 device: str = 'cuda'   # torch.device("cuda" if torch.cuda.is_available() else "cpu")
                 ):
        
        os.environ["CUDA_VISIBLE_DEVICES"] = cuda_visible_device
        self.DEVICE = device
        
        self.processor = WhisperProcessor.from_pretrained(model_name_or_path)
        self.tokenizer = self.processor.tokenizer

        self.config = WhisperConfig.from_pretrained(model_name_or_path)

        self.model = WhisperForConditionalGeneration(
                config=self.config
            ).from_pretrained(
                            pretrained_model_name_or_path = model_name_or_path, 
                            torch_dtype = self.config.torch_dtype,
                            # device_map=DEVICE,      # 'balanced', 'balanced_low_0', 'sequential', 'cuda', 'cpu'
                            low_cpu_mem_usage = True,
                        )
            
        # Move model to GPU
        if self.model.device.type != self.DEVICE:
            print(f'Moving model to {self.DEVICE}')
            self.model = self.model.to(self.DEVICE)
            self.model.eval()

        else:
            print(f'Model is already on {self.DEVICE}')
            self.model.eval()
            
        print('dtype of model acc to config: ', self.config.torch_dtype)
        print('dtype of loaded model: ', self.model.dtype)
        
    
    # audio = whisper.load_audio('test.wav')
    def load_audio(self, file: str, sr: int = SAMPLE_RATE, start_time: int = 0, dtype=np.float16):
        try:
            # This launches a subprocess to decode audio while down-mixing and resampling as necessary.
            # Requires the ffmpeg CLI and `ffmpeg-python` package to be installed.
            out, _ = (
                ffmpeg.input(file, ss=start_time, threads=0)
                .output("-", format="s16le", acodec="pcm_s16le", ac=1, ar=sr)
                .run(cmd=["ffmpeg", "-nostdin"], capture_stdout=True, capture_stderr=True)
            )
        except ffmpeg.Error as e:
            raise RuntimeError(f"Failed to load audio: {e.stderr.decode()}") from e

        # return np.frombuffer(out, np.int16).flatten().astype(np.float32) / 32768.0
        return np.frombuffer(out, np.int16).flatten().astype(dtype) / 32768.0


    # audio = whisper.pad_or_trim(audio)
    def _pad_or_trim(self, array, length: int = N_SAMPLES, *, axis: int = -1):
        """
        Pad or trim the audio array to N_SAMPLES, as expected by the encoder.
        """
        if torch.is_tensor(array):
            if array.shape[axis] > length:
                array = array.index_select(
                    dim=axis, index=torch.arange(length, device=array.device)
                )

            if array.shape[axis] < length:
                pad_widths = [(0, 0)] * array.ndim
                pad_widths[axis] = (0, length - array.shape[axis])
                array = F.pad(array, [pad for sizes in pad_widths[::-1] for pad in sizes])
        else:
            if array.shape[axis] > length:
                array = array.take(indices=range(length), axis=axis)

            if array.shape[axis] < length:
                pad_widths = [(0, 0)] * array.ndim
                pad_widths[axis] = (0, length - array.shape[axis])
                array = np.pad(array, pad_widths)

        return array
    
    def transcribe(self, audio: np.ndarray, language: str = "english"):
        # audio = load_audio(audio)
        audio = self._pad_or_trim(audio)
        input_features = self.processor(audio, sampling_rate=SAMPLE_RATE, return_tensors="pt").input_features.half().to(self.DEVICE)
        with torch.no_grad():
            predicted_ids = self.model.generate(
                input_features,
                num_beams = 1,
                language=language,
                task="transcribe",
                use_cache=True,
                is_multilingual=True,
                return_timestamps=True,
            )
        
        transcription = self.tokenizer.batch_decode(predicted_ids, skip_special_tokens=True)[0]
        return transcription.strip()