|
|
import torch |
|
|
from diffusers import ConfigMixin, ModelMixin |
|
|
from einops import rearrange |
|
|
from torch import nn |
|
|
import math |
|
|
|
|
|
|
|
|
class AudioProjModel(ModelMixin, ConfigMixin): |
|
|
def __init__( |
|
|
self, |
|
|
seq_len=5, |
|
|
blocks=12, |
|
|
channels=768, |
|
|
intermediate_dim=512, |
|
|
output_dim=768, |
|
|
context_tokens=32, |
|
|
): |
|
|
super().__init__() |
|
|
|
|
|
self.seq_len = seq_len |
|
|
self.blocks = blocks |
|
|
self.channels = channels |
|
|
self.input_dim = seq_len * blocks * channels |
|
|
self.intermediate_dim = intermediate_dim |
|
|
self.context_tokens = context_tokens |
|
|
self.output_dim = output_dim |
|
|
|
|
|
|
|
|
self.proj1 = nn.Linear(self.input_dim, intermediate_dim) |
|
|
self.proj2 = nn.Linear(intermediate_dim, intermediate_dim) |
|
|
self.proj3 = nn.Linear(intermediate_dim, context_tokens * output_dim) |
|
|
|
|
|
self.norm = nn.LayerNorm(output_dim) |
|
|
|
|
|
def forward(self, audio_embeds): |
|
|
if audio_embeds.dim() == 4: |
|
|
audio_embeds = audio_embeds.unsqueeze(0) |
|
|
video_length = audio_embeds.shape[1] |
|
|
audio_embeds = rearrange(audio_embeds, "bz f w b c -> (bz f) w b c") |
|
|
batch_size, window_size, blocks, channels = audio_embeds.shape |
|
|
audio_embeds = audio_embeds.view(batch_size, window_size * blocks * channels) |
|
|
|
|
|
audio_embeds = torch.relu(self.proj1(audio_embeds)) |
|
|
audio_embeds = torch.relu(self.proj2(audio_embeds)) |
|
|
|
|
|
context_tokens = self.proj3(audio_embeds).reshape(batch_size, self.context_tokens, self.output_dim) |
|
|
|
|
|
context_tokens = self.norm(context_tokens) |
|
|
context_tokens = rearrange(context_tokens, "(bz f) m c -> bz f m c", f=video_length) |
|
|
|
|
|
return context_tokens |
|
|
|
|
|
|
|
|
class PeriodicPositionalEncoding(nn.Module): |
|
|
def __init__(self, d_model, dropout=0.1, period=25, max_seq_len=600): |
|
|
super(PeriodicPositionalEncoding, self).__init__() |
|
|
self.dropout = nn.Dropout(p=dropout) |
|
|
pe = torch.zeros(period, d_model) |
|
|
position = torch.arange(0, period, dtype=torch.float).unsqueeze(1) |
|
|
div_term = torch.exp(torch.arange(0, d_model, 2).float() * (-math.log(10000.0) / d_model)) |
|
|
pe[:, 0::2] = torch.sin(position * div_term) |
|
|
pe[:, 1::2] = torch.cos(position * div_term) |
|
|
pe = pe.unsqueeze(0) |
|
|
repeat_num = (max_seq_len//period) + 1 |
|
|
pe = pe.repeat(1, repeat_num, 1) |
|
|
self.register_buffer('pe', pe) |
|
|
def forward(self, x): |
|
|
x = x + self.pe[:, :x.size(1), :] |
|
|
return self.dropout(x) |
|
|
|
|
|
|
|
|
if __name__ == "__main__": |
|
|
|
|
|
audio_proj = AudioProjModel( |
|
|
seq_len=5, |
|
|
blocks=12, |
|
|
channels=768, |
|
|
intermediate_dim=512, |
|
|
output_dim=768, |
|
|
context_tokens=32, |
|
|
) |
|
|
audio = torch.randn(1, 41, 5, 12, 768) |
|
|
|
|
|
output = audio_proj(audio) |
|
|
print(output.shape) |