Args: img_size (int): Input image size. patch_size (int): Patch size. in_chans (int): Number of input ima

Args: dim (int): Number of input channels. num_heads (int): Number of attention heads. qkv_bias (bool):

Args: kernel_size (Tuple): kernel size of the projection layer. stride (Tuple): stride of the projection layer.

Converts an id to a token, special tokens included

Converts a token to an id using the vocab, special tokens included

Initialize the weights.

Collects the state dict and dump to disk.

Converts a string in a sequence of tokens (string), using the tokenizer. Split in words for word-based vocabulary or sub-words for su

Tokenize a list of strings.

same output as PIL.ImageOps.autocontrast

same output as PIL.ImageEnhance.Contrast

same output as PIL.ImageEnhance.Color

same output as PIL.ImageEnhance.Contrast

Converts a sequence of tokens in a single string.

Setup the optimizer. We provide a reasonable default that works well. If you want to use something else, you can pass a tuple in the

same output as PIL.ImageOps.equalize PIL's implementation is different from cv2.equalize

Input shape: Batch x Time x Channel attention_mask: [bsz, q_len]

`input_ids_shape` is expected to be [bsz x seqlen].

Input shape: Batch x Time x Channel

Args: hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)` attention_mask (

r""" Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence toke

r""" Args: input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`): Indices of input sequence toke