a-One-Fan/DL-Art-School
1
0
Fork 0
forked from mrq/DL-Art-School
Code Pull Requests Activity

get rid of nil tokens in <2>

Browse Source
This commit is contained in:
James Betker 2022-01-27 22:45:57 -07:00
parent 0152174c0e
commit 935a4e853e
1 changed files with 1 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
codes/models/gpt_voice/unet_diffusion_tts2.py
View File

@ -135,7 +135,6 @@ class DiffusionTts(nn.Module):
scale_factor=2, scale_factor=2,
conditioning_inputs_provided=True, conditioning_inputs_provided=True,
time_embed_dim_multiplier=4, time_embed_dim_multiplier=4,
nil_guidance_fwd_proportion=.3,
): ):
super().__init__() super().__init__()
@ -154,8 +153,6 @@ class DiffusionTts(nn.Module):
self.num_head_channels = num_head_channels self.num_head_channels = num_head_channels
self.num_heads_upsample = num_heads_upsample self.num_heads_upsample = num_heads_upsample
self.dims = dims self.dims = dims
self.nil_guidance_fwd_proportion = nil_guidance_fwd_proportion
self.mask_token_id = num_tokens
padding = 1 if kernel_size == 3 else 2 padding = 1 if kernel_size == 3 else 2
@ -186,7 +183,7 @@ class DiffusionTts(nn.Module):
for level, (mult, num_blocks) in enumerate(zip(channel_mult, num_res_blocks)): for level, (mult, num_blocks) in enumerate(zip(channel_mult, num_res_blocks)):
if ds in token_conditioning_resolutions: if ds in token_conditioning_resolutions:
token_conditioning_block = nn.Embedding(num_tokens+1, ch) token_conditioning_block = nn.Embedding(num_tokens, ch)
token_conditioning_block.weight.data.normal_(mean=0.0, std=.02) token_conditioning_block.weight.data.normal_(mean=0.0, std=.02)
self.input_blocks.append(token_conditioning_block) self.input_blocks.append(token_conditioning_block)
token_conditioning_blocks.append(token_conditioning_block) token_conditioning_blocks.append(token_conditioning_block)
@ -289,23 +286,6 @@ class DiffusionTts(nn.Module):
zero_module(conv_nd(dims, model_channels, out_channels, kernel_size, padding=padding)), zero_module(conv_nd(dims, model_channels, out_channels, kernel_size, padding=padding)),
) )
def load_state_dict(self, state_dict: 'OrderedDict[str, Tensor]',
strict: bool = True):
# Temporary hack to allow the addition of nil-guidance token embeddings to the existing guidance embeddings.
lsd = self.state_dict()
revised = 0
for i, blk in enumerate(self.input_blocks):
if isinstance(blk, nn.Embedding):
key = f'input_blocks.{i}.weight'
if state_dict[key].shape[0] != lsd[key].shape[0]:
t = torch.randn_like(lsd[key]) * .02
t[:state_dict[key].shape[0]] = state_dict[key]
state_dict[key] = t
revised += 1
print(f"Loaded experimental unet_diffusion_net with {revised} modifications.")
return super().load_state_dict(state_dict, strict)
def forward(self, x, timesteps, tokens, conditioning_input=None): def forward(self, x, timesteps, tokens, conditioning_input=None):
""" """
@ -333,11 +313,6 @@ class DiffusionTts(nn.Module):
else: else:
emb = emb1 emb = emb1
# Mask out guidance tokens for un-guided diffusion.
if self.training and self.nil_guidance_fwd_proportion > 0:
token_mask = torch.rand(tokens.shape, device=tokens.device) < self.nil_guidance_fwd_proportion
tokens = torch.where(token_mask, self.mask_token_id, tokens)
h = x.type(self.dtype) h = x.type(self.dtype)
for k, module in enumerate(self.input_blocks): for k, module in enumerate(self.input_blocks):
if isinstance(module, nn.Embedding): if isinstance(module, nn.Embedding):