make sure forward sum loss actually runs, cleanup

Author: lucidrains

README.md

@@ -108,7 +108,7 @@ raw_audio = torch.randn(4, 327680)
 prompt = torch.randn(4, 32768)               # they randomly excised a range on the audio for the prompt during training, eventually will take care of this auto-magically
 
 text = torch.randint(0, 100, (4, 100))
-text_lens = torch.tensor([100, 50 , 80, 120])
+text_lens = torch.tensor([100, 50 , 80, 100])
 
 # forwards and backwards
 

naturalspeech2_pytorch/aligner.py

@@ -2,11 +2,14 @@
 import numpy as np
 
 import torch
-from torch import nn
+from torch import nn, Tensor
 from torch.nn import Module
 import torch.nn.functional as F
 
-from einops import rearrange
+from einops import rearrange, repeat
+
+from beartype import beartype
+from beartype.typing import Optional
 
 def exists(val):
     return val is not None
@@ -22,7 +25,6 @@ def __init__(
     ):
         super().__init__()
         self.temperature = temperature
-        self.softmax = torch.nn.Softmax(dim=3)
 
         self.key_layers = nn.ModuleList([
             nn.Conv1d(
@@ -50,7 +52,13 @@ def __init__(
             nn.Conv1d(dim_in, attn_channels, kernel_size=1, padding=0, bias=True)
         ])
 
-    def forward(self, queries: torch.Tensor, keys: torch.Tensor, mask: torch.Tensor = None):
+    @beartype
+    def forward(
+        self,
+        queries: Tensor,
+        keys: Tensor,
+        mask: Optional[Tensor] = None
+    ):
         key_out = keys
         for layer in self.key_layers:
             key_out = layer(key_out)
@@ -61,12 +69,15 @@ def forward(self, queries: torch.Tensor, keys: torch.Tensor, mask: torch.Tensor
 
         key_out = rearrange(key_out, 'b c t -> b t c')
         query_out = rearrange(query_out, 'b c t -> b t c')
-        attn_logp = torch.cdist(query_out, key_out).unsqueeze(1)
+
+        attn_logp = torch.cdist(query_out, key_out)
+        attn_logp = rearrange(attn_logp, 'b ... -> b 1 ...')
 
         if exists(mask):
-            attn_logp.data.masked_fill_(~mask.bool().unsqueeze(2), -float("inf"))
+            mask = rearrange(mask.bool(), '... c -> ... 1 c')
+            attn_logp.data.masked_fill_(~mask, -torch.finfo(attn_logp.dtype).max)
 
-        attn = self.softmax(attn_logp)
+        attn = attn_logp.softmax(dim = -1)
         return attn, attn_logp
 
 def pad_tensor(input, pad, value=0):
@@ -110,34 +121,38 @@ def maximum_path(value, mask, const=None):
     path = path.to(dtype=dtype)
     return path
 
-class ForwardSumLoss():
-    def __init__(self, blank_logprob=-1):
+class ForwardSumLoss(Module):
+    def __init__(
+        self,
+        blank_logprob = -1
+    ):
         super().__init__()
-        self.log_softmax = torch.nn.LogSoftmax(dim=-1)
-        self.ctc_loss = torch.nn.CTCLoss(zero_infinity=True)
         self.blank_logprob = blank_logprob
 
-    def forward(self, attn_logprob, in_lens, out_lens):
-        key_lens = in_lens
-        query_lens = out_lens
+        self.ctc_loss = torch.nn.CTCLoss(
+            blank = 0,  # check this value
+            zero_infinity = True
+        )
+
+    def forward(self, attn_logprob, key_lens, query_lens):
+        device, blank_logprob  = attn_logprob.device, self.blank_logprob
         max_key_len = attn_logprob.size(-1)
 
         # Reorder input to [query_len, batch_size, key_len]
-        attn_logprob = rearrange(attn_logprob, 'b c t -> c b t')
+        attn_logprob = rearrange(attn_logprob, 'b 1 c t -> c b t')
 
         # Add blank label
-        attn_logprob = F.pad(attn_logprob, (1, 0, 0, 0, 0, 0), self.blank_logprob)
+        attn_logprob = F.pad(attn_logprob, (1, 0, 0, 0, 0, 0), value = blank_logprob)
 
         # Convert to log probabilities
         # Note: Mask out probs beyond key_len
-        device = attn_logprob.device
         attn_logprob.masked_fill_(torch.arange(max_key_len + 1, device=device, dtype=torch.long).view(1, 1, -1) > key_lens.view(1, -1, 1), -1e15)
 
-        attn_logprob = self.log_softmax(attn_logprob)
+        attn_logprob = attn_logprob.log_softmax(dim = -1)
 
         # Target sequences
-        target_seqs = torch.arange(1, max_key_len + 1, device=device, dtype=torch.long).unsqueeze(0)
-        target_seqs = target_seqs.repeat(key_lens.numel(), 1)
+        target_seqs = torch.arange(1, max_key_len + 1, device=device, dtype=torch.long)
+        target_seqs = repeat(target_seqs, 'n -> b n', b = key_lens.numel())
 
         # Evaluate CTC loss
         cost = self.ctc_loss(attn_logprob, target_seqs, query_lens, key_lens)

naturalspeech2_pytorch/naturalspeech2_pytorch.py

@@ -1448,6 +1448,8 @@ def forward(
             if not exists(text_lens):
                 text_lens = torch.full((batch,), text_max_length, device = self.device, dtype = torch.long)
 
+            text_lens.clamp_(max = text_max_length)
+
             text_mask = rearrange(create_mask(text_lens, text_max_length), 'b n -> b 1 n')
 
             prompt = self.process_prompt(prompt)
@@ -1475,6 +1477,8 @@ def forward(
             if not exists(mel_lens):
                 mel_lens = torch.full((batch,), mel_max_length, device = self.device, dtype = torch.long)
 
+            mel_lens.clamp_(max = mel_max_length)
+
             mel_mask = rearrange(create_mask(mel_lens, mel_max_length), 'b n -> b 1 n')
 
             # alignment

naturalspeech2_pytorch/version.py

@@ -1 +1 @@
-__version__ = '0.1.0'
+__version__ = '0.1.1'