implemented grad norm penalty+moved logs to dataset
This commit is contained in:
SayanoAI
+86
-63
@@ -3,14 +3,11 @@ import json
|
||||
import os
|
||||
import shutil
|
||||
import traceback
|
||||
|
||||
from .lib.audio import SR_MAP
|
||||
from .lib.train import utils
|
||||
import datetime
|
||||
|
||||
from random import shuffle, randint
|
||||
|
||||
import torch
|
||||
|
||||
torch.backends.cudnn.deterministic = False
|
||||
torch.backends.cudnn.benchmark = False
|
||||
from torch.nn import functional as F
|
||||
@@ -30,7 +27,6 @@ from .lib.train.data_utils import (
|
||||
TextAudioCollate,
|
||||
DistributedBucketSampler,
|
||||
)
|
||||
|
||||
from .lib.train.losses import generator_loss, discriminator_loss, feature_loss, kl_loss
|
||||
from .lib.train.mel_processing import mel_spectrogram_torch, spec_to_mel_torch
|
||||
|
||||
@@ -122,7 +118,7 @@ def train_model(hps: "utils.HParams"):
|
||||
|
||||
def run(rank, n_gpus, hps, device):
|
||||
print(f"{__name__=}")
|
||||
global global_step, least_loss, loss_file
|
||||
global global_step, least_loss, loss_file, best_model_name
|
||||
global_step = 0
|
||||
loss_file = os.path.join(hps.model_dir,"losses.json")
|
||||
|
||||
@@ -130,7 +126,10 @@ def run(rank, n_gpus, hps, device):
|
||||
with open(loss_file,"r") as f:
|
||||
data: dict = json.load(f)
|
||||
least_loss = data.get("least_loss",40)
|
||||
else: least_loss = 40
|
||||
best_model_name = data.get("best_model_name","")
|
||||
else:
|
||||
least_loss = 40
|
||||
best_model_name = ""
|
||||
|
||||
if hps.version == "v1":
|
||||
from .lib.infer_pack.models import (
|
||||
@@ -304,16 +303,16 @@ def run(rank, n_gpus, hps, device):
|
||||
|
||||
|
||||
def train_and_evaluate(
|
||||
rank, epoch, hps, nets, optims, schedulers, scaler, loaders, logger, writers, cache
|
||||
rank, epoch, hps, nets, optims, _, scaler, loaders, logger, writers, cache
|
||||
):
|
||||
net_g, net_d = nets
|
||||
optim_g, optim_d = optims
|
||||
train_loader, eval_loader = loaders
|
||||
train_loader, _ = loaders
|
||||
if writers is not None:
|
||||
writer, writer_eval = writers
|
||||
writer, _ = writers
|
||||
|
||||
train_loader.batch_sampler.set_epoch(epoch)
|
||||
global global_step, least_loss, loss_file
|
||||
global global_step, least_loss, loss_file, best_model_name
|
||||
|
||||
net_g.train()
|
||||
net_d.train()
|
||||
@@ -479,13 +478,41 @@ def train_and_evaluate(
|
||||
) # slice
|
||||
|
||||
# Discriminator
|
||||
y_d_hat_r, y_d_hat_g, _, _ = net_d(wave, y_hat.detach())
|
||||
gen_wave = y_hat.detach()
|
||||
y_d_hat_r, y_d_hat_g, _, _ = net_d(wave, gen_wave)
|
||||
|
||||
gradient_penalty = torch.Tensor([0.]).to(wave.device)
|
||||
if hps.train.get("gradient_lambda",0.)>0:
|
||||
# Compute the gradient penalty
|
||||
# Randomly interpolate between real and generated data
|
||||
size = [1]*wave.ndim
|
||||
size[0] = wave.size(0)
|
||||
alpha = torch.rand(*size, device=wave.device)
|
||||
interpolated = alpha * wave + (1 - alpha) * gen_wave
|
||||
interpolated.requires_grad_(True)
|
||||
# Get the discriminator output for the interpolated data
|
||||
_, disc_interpolated_output, _, _ = net_d(wave, interpolated)
|
||||
# Compute gradients of discriminator output w.r.t. interpolated data
|
||||
for output in disc_interpolated_output:
|
||||
gradient = torch.autograd.grad(
|
||||
outputs=output,
|
||||
inputs=interpolated,
|
||||
grad_outputs=torch.ones(output.size(), device=wave.device),
|
||||
create_graph=True,
|
||||
only_inputs=True
|
||||
)[0]
|
||||
gradient = gradient.view(gradient.size(0), -1)
|
||||
gradient_penalty += ((gradient.norm(2,dim=-1) - 1) ** 2).mean()
|
||||
gradient_penalty = gradient_penalty.squeeze()
|
||||
|
||||
with autocast(enabled=False):
|
||||
loss_disc, losses_disc_r, losses_disc_g = discriminator_loss(
|
||||
y_d_hat_r, y_d_hat_g
|
||||
)
|
||||
if hps.train.get("gradient_lambda",0.)>0: gradient_penalty*=hps.train.gradient_lambda
|
||||
|
||||
optim_d.zero_grad()
|
||||
scaler.scale(loss_disc).backward()
|
||||
scaler.scale(loss_disc+gradient_penalty).backward()
|
||||
scaler.unscale_(optim_d)
|
||||
grad_norm_d = commons.clip_grad_value_(net_d.parameters(), None)
|
||||
scaler.step(optim_d)
|
||||
@@ -507,38 +534,35 @@ def train_and_evaluate(
|
||||
scaler.update()
|
||||
|
||||
if rank == 0:
|
||||
if loss_gen_all<least_loss:
|
||||
least_loss = loss_gen_all
|
||||
|
||||
if hps.save_best_model and epoch>min(hps.total_epoch*.2,20): #start saving after 20% or 20 epochs
|
||||
if hasattr(net_g, "module"):
|
||||
ckpt = net_g.module.state_dict()
|
||||
else:
|
||||
ckpt = net_g.state_dict()
|
||||
best_model_name = f"{hps.name}_e{epoch}_s{global_step}_loss{loss_gen_all:2.2f}"
|
||||
status = save_checkpoint(ckpt,best_model_name,epoch,hps)
|
||||
if epoch>=max(hps.total_epoch*.2,20): #start saving best model after 20% or 20 epochs
|
||||
if loss_gen_all+loss_disc<least_loss:
|
||||
least_loss = loss_gen_all+loss_disc
|
||||
logger.info(f"[lowest loss] {least_loss=:.3f}: {loss_disc=:.3f}, {loss_gen=:.3f}, {loss_fm=:.3f}, {loss_mel=:.3f}, {loss_kl=:.3f}")
|
||||
|
||||
if hps.save_best_model:
|
||||
if hasattr(net_g, "module"):
|
||||
ckpt = net_g.module.state_dict()
|
||||
else:
|
||||
ckpt = net_g.state_dict()
|
||||
best_model_name = f"{hps.name}_e{epoch}_s{global_step}_loss{least_loss:2.2f}"
|
||||
status = save_checkpoint(ckpt,best_model_name,epoch,hps)
|
||||
logger.info(f"=== saving best model {best_model_name}: {status=} ===")
|
||||
|
||||
with open(loss_file,"w") as f:
|
||||
json.dump(dict(least_loss=least_loss.item(),best_model=os.path.join(hps.model_dir,best_model_name+".pth")),f)
|
||||
logger.info(f"=== saving best model: epoch {hps.name}_e{epoch}_{loss_gen_all:2.2f}: {status=} ===")
|
||||
logger.info(f"[lowest loss] loss_disc={loss_disc:.3f}, loss_gen={loss_gen:.3f}, loss_fm={loss_fm:.3f},loss_mel={loss_mel:.3f}, loss_kl={loss_kl:.3f}")
|
||||
json.dump(dict(least_loss=least_loss.item(),best_model_name=best_model_name),f)
|
||||
|
||||
if hps.train.log_interval>0 and global_step % hps.train.log_interval == 0:
|
||||
lr = optim_g.param_groups[0]["lr"]
|
||||
logger.info(
|
||||
"Train Epoch: {} [{:.0f}%]".format(epoch, 100.0 * batch_idx / len(train_loader))
|
||||
)
|
||||
logger.info(f"Train Epoch: {epoch} [{100.0 * batch_idx / len(train_loader):.0f}%]")
|
||||
# Amor For Tensorboard display
|
||||
if loss_mel > 75:
|
||||
loss_mel = 75
|
||||
if loss_kl > 9:
|
||||
loss_kl = 9
|
||||
|
||||
logger.info(f"epoch={epoch} steps={global_step} lr={lr} total_loss={loss_gen_all}")
|
||||
logger.info(f"{epoch=} {global_step=} {lr=:.2E} {loss_disc=:.3f} {gradient_penalty.item()=:.2E}")
|
||||
logger.info(f"{loss_gen_all=:.3f}: {loss_gen=:.3f}, {loss_fm=:.3f}, {loss_mel=:.3f}, {loss_kl=:.3f}")
|
||||
|
||||
logger.info(
|
||||
f"loss_disc={loss_disc:.3f}, loss_gen={loss_gen:.3f}, loss_fm={loss_fm:.3f},loss_mel={loss_mel:.3f}, loss_kl={loss_kl:.3f}"
|
||||
)
|
||||
|
||||
scalar_dict = {
|
||||
"total/loss/all": loss_gen_all,
|
||||
"total/loss/g": loss_gen,
|
||||
@@ -547,35 +571,31 @@ def train_and_evaluate(
|
||||
"total/loss/mel": loss_mel,
|
||||
"total/loss/kl": loss_kl,
|
||||
"learning_rate": lr,
|
||||
"grad_norm_d": grad_norm_d,
|
||||
"grad_norm_g": grad_norm_g,
|
||||
"gradient/grad_norm_d": grad_norm_d,
|
||||
"gradient/grad_norm_g": grad_norm_g,
|
||||
"gradient/gradient_penalty": gradient_penalty,
|
||||
**{f"loss/g/{i}": v for i, v in enumerate(losses_gen)},
|
||||
**{f"loss/d_r/{i}": v for i, v in enumerate(losses_disc_r)},
|
||||
**{f"loss/d_g/{i}": v for i, v in enumerate(losses_disc_g)}
|
||||
}
|
||||
|
||||
scalar_dict.update(
|
||||
{"loss/g/{}".format(i): v for i, v in enumerate(losses_gen)}
|
||||
)
|
||||
scalar_dict.update(
|
||||
{"loss/d_r/{}".format(i): v for i, v in enumerate(losses_disc_r)}
|
||||
)
|
||||
scalar_dict.update(
|
||||
{"loss/d_g/{}".format(i): v for i, v in enumerate(losses_disc_g)}
|
||||
)
|
||||
image_dict = {
|
||||
"slice/mel_org": utils.plot_spectrogram_to_numpy(
|
||||
y_mel[0].data.cpu().numpy()
|
||||
),
|
||||
"slice/mel_gen": utils.plot_spectrogram_to_numpy(
|
||||
y_hat_mel[0].data.cpu().numpy()
|
||||
),
|
||||
"all/mel": utils.plot_spectrogram_to_numpy(
|
||||
mel[0].data.cpu().numpy()
|
||||
),
|
||||
"slice/mel_org": utils.plot_spectrogram_to_numpy(y_mel[0].data.cpu().numpy()),
|
||||
"slice/mel_gen": utils.plot_spectrogram_to_numpy(y_hat_mel[0].data.cpu().numpy()),
|
||||
"all/mel": utils.plot_spectrogram_to_numpy(mel[0].data.cpu().numpy()),
|
||||
}
|
||||
|
||||
audio_dict = {
|
||||
"slice/wave_org": wave.flatten(),
|
||||
"slice/wave_gen": y_hat.flatten()
|
||||
}
|
||||
utils.summarize(
|
||||
writer=writer,
|
||||
global_step=global_step,
|
||||
images=image_dict,
|
||||
scalars=scalar_dict,
|
||||
audios=audio_dict,
|
||||
audio_sampling_rate=SR_MAP[hps.sample_rate]
|
||||
)
|
||||
global_step += 1
|
||||
# /Run steps
|
||||
@@ -588,14 +608,14 @@ def train_and_evaluate(
|
||||
optim_g,
|
||||
hps.train.learning_rate,
|
||||
epoch,
|
||||
os.path.join(hps.model_dir, f"G_23333{hps.total_epoch}.pth"),
|
||||
os.path.join(hps.model_dir, "G_23333.pth"),
|
||||
)
|
||||
utils.save_checkpoint(
|
||||
net_d,
|
||||
optim_d,
|
||||
hps.train.learning_rate,
|
||||
epoch,
|
||||
os.path.join(hps.model_dir, f"D_23333{hps.total_epoch}.pth"),
|
||||
os.path.join(hps.model_dir, "D_23333.pth"),
|
||||
)
|
||||
else:
|
||||
utils.save_checkpoint(
|
||||
@@ -603,14 +623,14 @@ def train_and_evaluate(
|
||||
optim_g,
|
||||
hps.train.learning_rate,
|
||||
epoch,
|
||||
os.path.join(hps.model_dir, f"G_{global_step}.pth"),
|
||||
os.path.join(hps.model_dir, f"G_{epoch}.pth"),
|
||||
)
|
||||
utils.save_checkpoint(
|
||||
net_d,
|
||||
optim_d,
|
||||
hps.train.learning_rate,
|
||||
epoch,
|
||||
os.path.join(hps.model_dir, f"D_{global_step}.pth"),
|
||||
os.path.join(hps.model_dir, f"D_{epoch}.pth"),
|
||||
)
|
||||
if rank == 0 and hps.save_every_weights:
|
||||
if hasattr(net_g, "module"):
|
||||
@@ -621,7 +641,7 @@ def train_and_evaluate(
|
||||
logger.info(f"saving ckpt {hps.name}_e{epoch}: {status}")
|
||||
|
||||
if rank == 0:
|
||||
logger.info("====> Epoch: {} {}".format(epoch, epoch_recorder.record()))
|
||||
logger.info(f"====> Epoch {epoch}: {loss_gen_all=:.3f} {loss_disc=:.3f} {epoch_recorder.record()}")
|
||||
if epoch >= hps.total_epoch and rank == 0:
|
||||
logger.info("Training is done. The program is closed.")
|
||||
|
||||
@@ -632,12 +652,15 @@ def train_and_evaluate(
|
||||
if hps.save_best_model:
|
||||
with open(loss_file,"r") as f:
|
||||
data = json.load(f)
|
||||
best_model = data["best_model"]
|
||||
if os.path.isfile(best_model):
|
||||
shutil.copy(best_model,os.path.join(os.path.dirname(hps.model_path),f"{hps.name}_{hps.sample_rate}-best.pth"))
|
||||
best_model_name = data.get("best_model_name","")
|
||||
best_model_path = os.path.join(hps.model_dir,f"{best_model_name}.pth")
|
||||
if os.path.isfile(best_model_path):
|
||||
shutil.copy(best_model_path,os.path.join(
|
||||
os.path.dirname(hps.model_path),
|
||||
f"{os.path.basename(hps.model_path).split('.')[0]}-best.pth"))
|
||||
|
||||
status = save_checkpoint(ckpt,hps.name,epoch,hps,model_path=hps.model_path)
|
||||
logger.info(f"saving final ckpt: {status}")
|
||||
logger.info(f"saving final ckpt {hps.model_path}: {status}")
|
||||
sleep(1)
|
||||
os._exit(0)
|
||||
|
||||
|
||||
Reference in New Issue
Block a user