mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-16 20:52:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

minor updates

Browse Source
This commit is contained in:
zr_jin 2024-10-22 12:53:44 +08:00
parent ca3b495c4f
commit 3c3db1ae69
3 changed files with 20 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
egs/libritts/TTS/vits/infer.py
View File

@ -152,7 +152,7 @@ def infer_dataset(
audio_lens = batch["audio_lens"].tolist() audio_lens = batch["audio_lens"].tolist()
cut_ids = [cut.id for cut in batch["cut"]] cut_ids = [cut.id for cut in batch["cut"]]
sids = ["_".join(cut_id.split("_")[:2]) for cut_id in cut_ids] sids = ["_".join(cut_id.split("_")[:2]) for cut_id in cut_ids]
speakers = ( spembs = (
torch.Tensor(np.array([speaker_map.read(sid) for sid in sids])) torch.Tensor(np.array([speaker_map.read(sid) for sid in sids]))
.squeeze(1) .squeeze(1)
.to(device) .to(device)
@ -161,7 +161,7 @@ def infer_dataset(
audio_pred, _, durations = model.inference_batch( audio_pred, _, durations = model.inference_batch(
text=tokens, text=tokens,
text_lengths=tokens_lens, text_lengths=tokens_lens,
spembs=speakers, spembs=spembs,
) )
audio_pred = audio_pred.detach().cpu() audio_pred = audio_pred.detach().cpu()
# convert to samples # convert to samples

24
egs/libritts/TTS/vits/train.py
View File

@ -344,7 +344,7 @@ def prepare_input(
audio_lens = batch["audio_lens"].to(device) audio_lens = batch["audio_lens"].to(device)
features_lens = batch["features_lens"].to(device) features_lens = batch["features_lens"].to(device)
tokens = batch["tokens"] tokens = batch["tokens"]
speakers = ( spembs = (
torch.Tensor(np.array([speaker_map.read(sid) for sid in parse_sids(batch)])) torch.Tensor(np.array([speaker_map.read(sid) for sid in parse_sids(batch)]))
.squeeze(1) .squeeze(1)
.to(device) .to(device)
@ -361,7 +361,7 @@ def prepare_input(
# a tensor of shape (B, T) # a tensor of shape (B, T)
tokens = tokens.pad(mode="constant", padding_value=tokenizer.pad_id) tokens = tokens.pad(mode="constant", padding_value=tokenizer.pad_id)
return audio, audio_lens, features, features_lens, tokens, tokens_lens, speakers return audio, audio_lens, features, features_lens, tokens, tokens_lens, spembs
def train_one_epoch( def train_one_epoch(
@ -449,7 +449,7 @@ def train_one_epoch(
features_lens, features_lens,
tokens, tokens,
tokens_lens, tokens_lens,
speakers, spembs,
) = prepare_input(batch, tokenizer, device, train_speaker_map) ) = prepare_input(batch, tokenizer, device, train_speaker_map)
loss_info = MetricsTracker() loss_info = MetricsTracker()
@ -465,7 +465,7 @@ def train_one_epoch(
feats_lengths=features_lens, feats_lengths=features_lens,
speech=audio, speech=audio,
speech_lengths=audio_lens, speech_lengths=audio_lens,
spembs=speakers, spembs=spembs,
forward_generator=False, forward_generator=False,
) )
for k, v in stats_d.items(): for k, v in stats_d.items():
@ -484,7 +484,7 @@ def train_one_epoch(
feats_lengths=features_lens, feats_lengths=features_lens,
speech=audio, speech=audio,
speech_lengths=audio_lens, speech_lengths=audio_lens,
spembs=speakers, spembs=spembs,
forward_generator=True, forward_generator=True,
return_sample=params.batch_idx_train % params.log_interval == 0, return_sample=params.batch_idx_train % params.log_interval == 0,
) )
@ -651,7 +651,7 @@ def compute_validation_loss(
features_lens, features_lens,
tokens, tokens,
tokens_lens, tokens_lens,
speakers, spembs,
) = prepare_input(batch, tokenizer, device, dev_speaker_map) ) = prepare_input(batch, tokenizer, device, dev_speaker_map)
loss_info = MetricsTracker() loss_info = MetricsTracker()
@ -665,7 +665,7 @@ def compute_validation_loss(
feats_lengths=features_lens, feats_lengths=features_lens,
speech=audio, speech=audio,
speech_lengths=audio_lens, speech_lengths=audio_lens,
spembs=speakers, spembs=spembs,
forward_generator=False, forward_generator=False,
) )
assert loss_d.requires_grad is False assert loss_d.requires_grad is False
@ -680,7 +680,7 @@ def compute_validation_loss(
feats_lengths=features_lens, feats_lengths=features_lens,
speech=audio, speech=audio,
speech_lengths=audio_lens, speech_lengths=audio_lens,
spembs=speakers, spembs=spembs,
forward_generator=True, forward_generator=True,
) )
assert loss_g.requires_grad is False assert loss_g.requires_grad is False
@ -695,7 +695,7 @@ def compute_validation_loss(
inner_model = model.module if isinstance(model, DDP) else model inner_model = model.module if isinstance(model, DDP) else model
audio_pred, _, duration = inner_model.inference( audio_pred, _, duration = inner_model.inference(
text=tokens[0, : tokens_lens[0].item()], text=tokens[0, : tokens_lens[0].item()],
spembs=speakers[0], spembs=spembs[0],
) )
audio_pred = audio_pred.data.cpu().numpy() audio_pred = audio_pred.data.cpu().numpy()
audio_len_pred = ( audio_len_pred = (
@ -744,7 +744,7 @@ def scan_pessimistic_batches_for_oom(
features_lens, features_lens,
tokens, tokens,
tokens_lens, tokens_lens,
speakers, spembs,
) = prepare_input(batch, tokenizer, device, train_speaker_map) ) = prepare_input(batch, tokenizer, device, train_speaker_map)
try: try:
# for discriminator # for discriminator
@ -756,7 +756,7 @@ def scan_pessimistic_batches_for_oom(
feats_lengths=features_lens, feats_lengths=features_lens,
speech=audio, speech=audio,
speech_lengths=audio_lens, speech_lengths=audio_lens,
spembs=speakers, spembs=spembs,
forward_generator=False, forward_generator=False,
) )
optimizer_d.zero_grad() optimizer_d.zero_grad()
@ -770,7 +770,7 @@ def scan_pessimistic_batches_for_oom(
feats_lengths=features_lens, feats_lengths=features_lens,
speech=audio, speech=audio,
speech_lengths=audio_lens, speech_lengths=audio_lens,
spembs=speakers, spembs=spembs,
forward_generator=True, forward_generator=True,
) )
optimizer_g.zero_grad() optimizer_g.zero_grad()

7
egs/ljspeech/TTS/vits/generator.py
View File

@ -409,7 +409,12 @@ class VITSGenerator(torch.nn.Module):
g = self.global_emb(sids.view(-1)).unsqueeze(-1) g = self.global_emb(sids.view(-1)).unsqueeze(-1)
if self.spk_embed_dim is not None: if self.spk_embed_dim is not None:
# (B, global_channels, 1) # (B, global_channels, 1)
g_ = self.spemb_proj(F.normalize(spembs.unsqueeze(0))).unsqueeze(-1) if spembs.ndim == 2:
g_ = self.spemb_proj(F.normalize(spembs)).unsqueeze(-1)
elif spembs.ndim == 1:
g_ = self.spemb_proj(F.normalize(spembs.unsqueeze(0))).unsqueeze(-1)
else:
raise ValueError("spembs should be 1D or 2D (batch mode) tensor.")
if g is None: if g is None:
g = g_ g = g_
else: else: