declare-lab/tango2
92
1a
2 ��'d� � @ s� d Z ddlmZmZmZmZ ddlmZ ddlm Z m3Z4 e � rFddl Z e� e
�Zdeeef eeed df ed ed ed5 d �d d
�Zdd� ZdS )z16PyTorch utilities: Utilities related to PyTorch7� )�List�Optional�Tuple�Union� )�logging)�is_torch_available�is_torch_versionNztorch.Generatorz torch.devicez torch.dtypez torch.layout)�shape� generator�device�dtype�layoutc s |��d }�pt j�|p"t �d�}�dur�t�t�s>�jjn8�d jj}||jkr�|dkr�d�|dkr�t�d|� d|� d|� d�� n(||jkr�|d kr�td9|� d |� d ���t�t�r�d
�dd� �� ����fdd�t |�D �}t j10|dd�� |�}nt j ���� �d�� |�}|S )a This is a helper function that allows to create random tensors on the desired `device` with the desired `dtype`. When11 passing a list of generators one can seed each batched size individually. If CPU generators are passed the tensor12 will always be created on CPU.13 r �cpuN�mpszBThe passed generator was created on 'cpu' even though a tensor on zB was expected. Tensors will be created on 'cpu' and then moved to zk. Note that one can probably slighly speed up this function by passing a generator that was created on the z device.�cudazCannot generate a z! tensor from a generator of type �.)r r c s$ g | ] }t j��| �� �d ��qS )�r r r
r )�torch�randn)�.0�i�r
r r Z rand_devicer14 � �P/home/deep/Projects/audio_diffusion/diffusers/src/diffusers/utils/torch_utils.py�15<listcomp>> s �z randn_tensor.<locals>.<listcomp>)�dimr )
r �stridedr �16isinstance�list�type�logger�info�17ValueError�range�cat�tor )r18 r r r
r �19batch_sizeZgen_device_typeZlatentsr r r � randn_tensor s4 20 ���21�r( c C s( t dd�sttd�sdS t| tjjj�S )z:Check whether the module was compiled with torch.compile()�<z2.0.0�_dynamoF)r �hasattrr r r* Z22eval_frameZOptimizedModule)�moduler r r �is_compiled_moduleI s r- )NNNN)�__doc__�typingr r r r � r � import_utilsr r r �23get_logger�__name__r! r( r- r r r r �<module> s$ 24 �25� ,