mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-26 18:24:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #1 from luomingshuang/lossrecord-v2

Browse Source 
style check with flake8 and black
This commit is contained in:
Mingshuang Luo 2021-09-29 19:29:25 +08:00 committed by GitHub
commit b35eed961a
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 63 additions and 163 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

63
egs/librispeech/ASR/conformer_ctc/train.py
View File

@ -59,10 +59,7 @@ def get_parser():
) )
parser.add_argument( parser.add_argument(
"--world-size", "--world-size", type=int, default=1, help="Number of GPUs for DDP training.",
type=int,
default=1,
help="Number of GPUs for DDP training.",
) )
parser.add_argument( parser.add_argument(
@ -80,10 +77,7 @@ def get_parser():
) )
parser.add_argument( parser.add_argument(
"--num-epochs", "--num-epochs", type=int, default=35, help="Number of epochs to train.",
type=int,
default=35,
help="Number of epochs to train.",
) )
parser.add_argument( parser.add_argument(
@ -230,10 +224,7 @@ def load_checkpoint_if_available(
filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt" filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt"
saved_params = load_checkpoint( saved_params = load_checkpoint(
filename, filename, model=model, optimizer=optimizer, scheduler=scheduler,
model=model,
optimizer=optimizer,
scheduler=scheduler,
) )
keys = [ keys = [
@ -335,9 +326,7 @@ def compute_loss(
decoding_graph = graph_compiler.compile(token_ids) decoding_graph = graph_compiler.compile(token_ids)
dense_fsa_vec = k2.DenseFsaVec( dense_fsa_vec = k2.DenseFsaVec(
nnet_output, nnet_output, supervision_segments, allow_truncate=params.subsampling_factor - 1,
supervision_segments,
allow_truncate=params.subsampling_factor - 1,
) )
ctc_loss = k2.ctc_loss( ctc_loss = k2.ctc_loss(
@ -374,12 +363,12 @@ def compute_loss(
assert loss.requires_grad == is_training assert loss.requires_grad == is_training
info = LossRecord() info = LossRecord()
info['frames'] = supervision_segments[:, 2].sum().item() info["frames"] = supervision_segments[:, 2].sum().item()
info['ctc_loss'] = ctc_loss.detach().cpu().item() info["ctc_loss"] = ctc_loss.detach().cpu().item()
if params.att_rate != 0.0: if params.att_rate != 0.0:
info['att_loss'] = att_loss.detach().cpu().item() info["att_loss"] = att_loss.detach().cpu().item()
info['loss'] = loss.detach().cpu().item() info["loss"] = loss.detach().cpu().item()
return loss, info return loss, info
@ -410,7 +399,7 @@ def compute_validation_loss(
if world_size > 1: if world_size > 1:
tot_loss.reduce(loss.device) tot_loss.reduce(loss.device)
loss_value = tot_loss['loss'] / tot_loss['frames'] loss_value = tot_loss["loss"] / tot_loss["frames"]
if loss_value < params.best_valid_loss: if loss_value < params.best_valid_loss:
params.best_valid_epoch = params.cur_epoch params.best_valid_epoch = params.cur_epoch
params.best_valid_loss = loss_value params.best_valid_loss = loss_value
@ -489,15 +478,9 @@ def train_one_epoch(
if tb_writer is not None: if tb_writer is not None:
loss_info.write_summary( loss_info.write_summary(
tb_writer, tb_writer, "train/current_", params.batch_idx_train
"train/current_",
params.batch_idx_train
)
tot_loss.write_summary(
tb_writer,
"train/tot_",
params.batch_idx_train
) )
tot_loss.write_summary(tb_writer, "train/tot_", params.batch_idx_train)
if batch_idx > 0 and batch_idx % params.valid_interval == 0: if batch_idx > 0 and batch_idx % params.valid_interval == 0:
logging.info("Computing validation loss") logging.info("Computing validation loss")
@ -509,17 +492,13 @@ def train_one_epoch(
world_size=world_size, world_size=world_size,
) )
model.train() model.train()
logging.info( logging.info(f"Epoch {params.cur_epoch}, validation: {valid_info}")
f"Epoch {params.cur_epoch}, validation: {valid_info}"
)
if tb_writer is not None: if tb_writer is not None:
valid_info.write_summary( valid_info.write_summary(
tb_writer, tb_writer, "train/valid_", params.batch_idx_train
"train/valid_",
params.batch_idx_train
) )
loss_value = tot_loss['loss'] / tot_loss['frames'] loss_value = tot_loss["loss"] / tot_loss["frames"]
params.train_loss = loss_value params.train_loss = loss_value
if params.train_loss < params.best_train_loss: if params.train_loss < params.best_train_loss:
params.best_train_epoch = params.cur_epoch params.best_train_epoch = params.cur_epoch
@ -563,10 +542,7 @@ def run(rank, world_size, args):
device = torch.device("cuda", rank) device = torch.device("cuda", rank)
graph_compiler = BpeCtcTrainingGraphCompiler( graph_compiler = BpeCtcTrainingGraphCompiler(
params.lang_dir, params.lang_dir, device=device, sos_token="<sos/eos>", eos_token="<sos/eos>",
device=device,
sos_token="<sos/eos>",
eos_token="<sos/eos>",
) )
logging.info("About to create model") logging.info("About to create model")
@ -607,9 +583,7 @@ def run(rank, world_size, args):
cur_lr = optimizer._rate cur_lr = optimizer._rate
if tb_writer is not None: if tb_writer is not None:
tb_writer.add_scalar( tb_writer.add_scalar("train/learning_rate", cur_lr, params.batch_idx_train)
"train/learning_rate", cur_lr, params.batch_idx_train
)
tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train) tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train)
if rank == 0: if rank == 0:
@ -629,10 +603,7 @@ def run(rank, world_size, args):
) )
save_checkpoint( save_checkpoint(
params=params, params=params, model=model, optimizer=optimizer, rank=rank,
model=model,
optimizer=optimizer,
rank=rank,
) )
logging.info("Done!") logging.info("Done!")

53
egs/librispeech/ASR/tdnn_lstm_ctc/train.py
View File

@ -58,10 +58,7 @@ def get_parser():
) )
parser.add_argument( parser.add_argument(
"--world-size", "--world-size", type=int, default=1, help="Number of GPUs for DDP training.",
type=int,
default=1,
help="Number of GPUs for DDP training.",
) )
parser.add_argument( parser.add_argument(
@ -79,10 +76,7 @@ def get_parser():
) )
parser.add_argument( parser.add_argument(
"--num-epochs", "--num-epochs", type=int, default=20, help="Number of epochs to train.",
type=int,
default=20,
help="Number of epochs to train.",
) )
parser.add_argument( parser.add_argument(
@ -209,10 +203,7 @@ def load_checkpoint_if_available(
filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt" filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt"
saved_params = load_checkpoint( saved_params = load_checkpoint(
filename, filename, model=model, optimizer=optimizer, scheduler=scheduler,
model=model,
optimizer=optimizer,
scheduler=scheduler,
) )
keys = [ keys = [
@ -312,9 +303,7 @@ def compute_loss(
decoding_graph = graph_compiler.compile(texts) decoding_graph = graph_compiler.compile(texts)
dense_fsa_vec = k2.DenseFsaVec( dense_fsa_vec = k2.DenseFsaVec(
nnet_output, nnet_output, supervision_segments, allow_truncate=params.subsampling_factor - 1,
supervision_segments,
allow_truncate=params.subsampling_factor - 1,
) )
loss = k2.ctc_loss( loss = k2.ctc_loss(
@ -328,8 +317,8 @@ def compute_loss(
assert loss.requires_grad == is_training assert loss.requires_grad == is_training
info = LossRecord() info = LossRecord()
info['frames'] = supervision_segments[:, 2].sum().item() info["frames"] = supervision_segments[:, 2].sum().item()
info['loss'] = loss.detach().cpu().item() info["loss"] = loss.detach().cpu().item()
return loss, info return loss, info
@ -363,7 +352,7 @@ def compute_validation_loss(
if world_size > 1: if world_size > 1:
tot_loss.reduce(loss.device) tot_loss.reduce(loss.device)
loss_value = tot_loss['loss'] / tot_loss['frames'] loss_value = tot_loss["loss"] / tot_loss["frames"]
if loss_value < params.best_valid_loss: if loss_value < params.best_valid_loss:
params.best_valid_epoch = params.cur_epoch params.best_valid_epoch = params.cur_epoch
@ -439,15 +428,9 @@ def train_one_epoch(
if tb_writer is not None: if tb_writer is not None:
loss_info.write_summary( loss_info.write_summary(
tb_writer, tb_writer, "train/current_", params.batch_idx_train
"train/current_",
params.batch_idx_train
)
tot_loss.write_summary(
tb_writer,
"train/tot_",
params.batch_idx_train
) )
tot_loss.write_summary(tb_writer, "train/tot_", params.batch_idx_train)
if batch_idx > 0 and batch_idx % params.valid_interval == 0: if batch_idx > 0 and batch_idx % params.valid_interval == 0:
valid_info = compute_validation_loss( valid_info = compute_validation_loss(
@ -458,17 +441,13 @@ def train_one_epoch(
world_size=world_size, world_size=world_size,
) )
model.train() model.train()
logging.info( logging.info(f"Epoch {params.cur_epoch}, validation {valid_info}")
f"Epoch {params.cur_epoch}, validation {valid_info}"
)
if tb_writer is not None: if tb_writer is not None:
valid_info.write_summary( valid_info.write_summary(
tb_writer, tb_writer, "train/valid_", params.batch_idx_train,
"train/valid_",
params.batch_idx_train,
) )
loss_value = tot_loss['loss'] / tot_loss['frames'] loss_value = tot_loss["loss"] / tot_loss["frames"]
params.train_loss = loss_value params.train_loss = loss_value
if params.train_loss < params.best_train_loss: if params.train_loss < params.best_train_loss:
@ -526,9 +505,7 @@ def run(rank, world_size, args):
model = DDP(model, device_ids=[rank]) model = DDP(model, device_ids=[rank])
optimizer = optim.AdamW( optimizer = optim.AdamW(
model.parameters(), model.parameters(), lr=params.lr, weight_decay=params.weight_decay,
lr=params.lr,
weight_decay=params.weight_decay,
) )
scheduler = StepLR(optimizer, step_size=8, gamma=0.1) scheduler = StepLR(optimizer, step_size=8, gamma=0.1)
@ -548,9 +525,7 @@ def run(rank, world_size, args):
if tb_writer is not None: if tb_writer is not None:
tb_writer.add_scalar( tb_writer.add_scalar(
"train/lr", "train/lr", scheduler.get_last_lr()[0], params.batch_idx_train,
scheduler.get_last_lr()[0],
params.batch_idx_train,
) )
tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train) tb_writer.add_scalar("train/epoch", epoch, params.batch_idx_train)

59
egs/yesno/ASR/tdnn/train.py
View File

@ -33,10 +33,7 @@ def get_parser():
) )
parser.add_argument( parser.add_argument(
"--world-size", "--world-size", type=int, default=1, help="Number of GPUs for DDP training.",
type=int,
default=1,
help="Number of GPUs for DDP training.",
) )
parser.add_argument( parser.add_argument(
@ -54,10 +51,7 @@ def get_parser():
) )
parser.add_argument( parser.add_argument(
"--num-epochs", "--num-epochs", type=int, default=15, help="Number of epochs to train.",
type=int,
default=15,
help="Number of epochs to train.",
) )
parser.add_argument( parser.add_argument(
@ -187,10 +181,7 @@ def load_checkpoint_if_available(
filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt" filename = params.exp_dir / f"epoch-{params.start_epoch-1}.pt"
saved_params = load_checkpoint( saved_params = load_checkpoint(
filename, filename, model=model, optimizer=optimizer, scheduler=scheduler,
model=model,
optimizer=optimizer,
scheduler=scheduler,
) )
keys = [ keys = [
@ -287,16 +278,12 @@ def compute_loss(
batch_size = nnet_output.shape[0] batch_size = nnet_output.shape[0]
supervision_segments = torch.tensor( supervision_segments = torch.tensor(
[[i, 0, nnet_output.shape[1]] for i in range(batch_size)], [[i, 0, nnet_output.shape[1]] for i in range(batch_size)], dtype=torch.int32,
dtype=torch.int32,
) )
decoding_graph = graph_compiler.compile(texts) decoding_graph = graph_compiler.compile(texts)
dense_fsa_vec = k2.DenseFsaVec( dense_fsa_vec = k2.DenseFsaVec(nnet_output, supervision_segments,)
nnet_output,
supervision_segments,
)
loss = k2.ctc_loss( loss = k2.ctc_loss(
decoding_graph=decoding_graph, decoding_graph=decoding_graph,
@ -309,8 +296,8 @@ def compute_loss(
assert loss.requires_grad == is_training assert loss.requires_grad == is_training
info = LossRecord() info = LossRecord()
info['frames'] = supervision_segments[:, 2].sum().item() info["frames"] = supervision_segments[:, 2].sum().item()
info['loss'] = loss.detach().cpu().item() info["loss"] = loss.detach().cpu().item()
return loss, info return loss, info
@ -344,7 +331,7 @@ def compute_validation_loss(
if world_size > 1: if world_size > 1:
tot_loss.reduce(loss.device) tot_loss.reduce(loss.device)
loss_value = tot_loss['loss'] / tot_loss['frames'] loss_value = tot_loss["loss"] / tot_loss["frames"]
if loss_value < params.best_valid_loss: if loss_value < params.best_valid_loss:
params.best_valid_epoch = params.cur_epoch params.best_valid_epoch = params.cur_epoch
@ -420,15 +407,9 @@ def train_one_epoch(
if tb_writer is not None: if tb_writer is not None:
loss_info.write_summary( loss_info.write_summary(
tb_writer, tb_writer, "train/current_", params.batch_idx_train
"train/current_",
params.batch_idx_train
)
tot_loss.write_summary(
tb_writer,
"train/tot_",
params.batch_idx_train
) )
tot_loss.write_summary(tb_writer, "train/tot_", params.batch_idx_train)
if batch_idx > 0 and batch_idx % params.valid_interval == 0: if batch_idx > 0 and batch_idx % params.valid_interval == 0:
valid_info = compute_validation_loss( valid_info = compute_validation_loss(
@ -439,17 +420,13 @@ def train_one_epoch(
world_size=world_size, world_size=world_size,
) )
model.train() model.train()
logging.info( logging.info(f"Epoch {params.cur_epoch}, validation {valid_info}")
f"Epoch {params.cur_epoch}, validation {valid_info}"
)
if tb_writer is not None: if tb_writer is not None:
valid_info.write_summary( valid_info.write_summary(
tb_writer, tb_writer, "train/valid_", params.batch_idx_train,
"train/valid_",
params.batch_idx_train,
) )
loss_value = tot_loss['loss'] / tot_loss['frames'] loss_value = tot_loss["loss"] / tot_loss["frames"]
params.train_loss = loss_value params.train_loss = loss_value
if params.train_loss < params.best_train_loss: if params.train_loss < params.best_train_loss:
@ -506,9 +483,7 @@ def run(rank, world_size, args):
model = DDP(model, device_ids=[rank]) model = DDP(model, device_ids=[rank])
optimizer = optim.SGD( optimizer = optim.SGD(
model.parameters(), model.parameters(), lr=params.lr, weight_decay=params.weight_decay,
lr=params.lr,
weight_decay=params.weight_decay,
) )
if checkpoints: if checkpoints:
@ -542,11 +517,7 @@ def run(rank, world_size, args):
) )
save_checkpoint( save_checkpoint(
params=params, params=params, model=model, optimizer=optimizer, scheduler=None, rank=rank,
model=model,
optimizer=optimizer,
scheduler=None,
rank=rank,
) )
logging.info("Done!") logging.info("Done!")

51
icefall/utils.py
View File

@ -107,9 +107,7 @@ def setup_logger(
formatter = f"%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] ({rank}/{world_size}) %(message)s" # noqa formatter = f"%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] ({rank}/{world_size}) %(message)s" # noqa
log_filename = f"{log_filename}-{date_time}-{rank}" log_filename = f"{log_filename}-{date_time}-{rank}"
else: else:
formatter = ( formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
"%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
)
log_filename = f"{log_filename}-{date_time}" log_filename = f"{log_filename}-{date_time}"
os.makedirs(os.path.dirname(log_filename), exist_ok=True) os.makedirs(os.path.dirname(log_filename), exist_ok=True)
@ -236,9 +234,7 @@ def get_texts(
return aux_labels.tolist() return aux_labels.tolist()
def store_transcripts( def store_transcripts(filename: Pathlike, texts: Iterable[Tuple[str, str]]) -> None:
filename: Pathlike, texts: Iterable[Tuple[str, str]]
) -> None:
"""Save predicted results and reference transcripts to a file. """Save predicted results and reference transcripts to a file.
Args: Args:
@ -369,19 +365,14 @@ def write_error_stats(
] ]
ali = list(filter(lambda x: x != [[], []], ali)) ali = list(filter(lambda x: x != [[], []], ali))
ali = [ ali = [
[ [ERR if x == [] else " ".join(x), ERR if y == [] else " ".join(y),]
ERR if x == [] else " ".join(x),
ERR if y == [] else " ".join(y),
]
for x, y in ali for x, y in ali
] ]
print( print(
" ".join( " ".join(
( (
ref_word ref_word if ref_word == hyp_word else f"({ref_word}->{hyp_word})"
if ref_word == hyp_word
else f"({ref_word}->{hyp_word})"
for ref_word, hyp_word in ali for ref_word, hyp_word in ali
) )
), ),
@ -391,9 +382,7 @@ def write_error_stats(
print("", file=f) print("", file=f)
print("SUBSTITUTIONS: count ref -> hyp", file=f) print("SUBSTITUTIONS: count ref -> hyp", file=f)
for count, (ref, hyp) in sorted( for count, (ref, hyp) in sorted([(v, k) for k, v in subs.items()], reverse=True):
[(v, k) for k, v in subs.items()], reverse=True
):
print(f"{count} {ref} -> {hyp}", file=f) print(f"{count} {ref} -> {hyp}", file=f)
print("", file=f) print("", file=f)
@ -407,9 +396,7 @@ def write_error_stats(
print(f"{count} {hyp}", file=f) print(f"{count} {hyp}", file=f)
print("", file=f) print("", file=f)
print( print("PER-WORD STATS: word corr tot_errs count_in_ref count_in_hyp", file=f)
"PER-WORD STATS: word corr tot_errs count_in_ref count_in_hyp", file=f
)
for _, word, counts in sorted( for _, word, counts in sorted(
[(sum(v[1:]), k, v) for k, v in words.items()], reverse=True [(sum(v[1:]), k, v) for k, v in words.items()], reverse=True
): ):
@ -428,7 +415,7 @@ class LossRecord(collections.defaultdict):
# makes undefined items default to int() which is zero. # makes undefined items default to int() which is zero.
super(LossRecord, self).__init__(int) super(LossRecord, self).__init__(int)
def __add__(self, other: 'LossRecord') -> 'LossRecord': def __add__(self, other: "LossRecord") -> "LossRecord":
ans = LossRecord() ans = LossRecord()
for k, v in self.items(): for k, v in self.items():
ans[k] = v ans[k] = v
@ -436,19 +423,19 @@ class LossRecord(collections.defaultdict):
ans[k] = ans[k] + v ans[k] = ans[k] + v
return ans return ans
def __mul__(self, alpha: float) -> 'LossRecord': def __mul__(self, alpha: float) -> "LossRecord":
ans = LossRecord() ans = LossRecord()
for k, v in self.items(): for k, v in self.items():
ans[k] = v * alpha ans[k] = v * alpha
return ans return ans
def __str__(self) -> str: def __str__(self) -> str:
ans = '' ans = ""
for k, v in self.norm_items(): for k, v in self.norm_items():
norm_value = '%.4g' % v norm_value = "%.4g" % v
ans += (str(k) + '=' + str(norm_value) + ', ') ans += str(k) + "=" + str(norm_value) + ", "
frames = str(self['frames']) frames = str(self["frames"])
ans += 'over ' + frames + ' frames.' ans += "over " + frames + " frames."
return ans return ans
def norm_items(self) -> List[Tuple[str, float]]: def norm_items(self) -> List[Tuple[str, float]]:
@ -456,10 +443,10 @@ class LossRecord(collections.defaultdict):
Returns a list of pairs, like: Returns a list of pairs, like:
[('ctc_loss', 0.1), ('att_loss', 0.07)] [('ctc_loss', 0.1), ('att_loss', 0.07)]
""" """
num_frames = self['frames'] if 'frames' in self else 1 num_frames = self["frames"] if "frames" in self else 1
ans = [] ans = []
for k, v in self.items(): for k, v in self.items():
if k != 'frames': if k != "frames":
norm_value = float(v) / num_frames norm_value = float(v) / num_frames
ans.append((k, norm_value)) ans.append((k, norm_value))
return ans return ans
@ -470,17 +457,13 @@ class LossRecord(collections.defaultdict):
all processes get the total. all processes get the total.
""" """
keys = sorted(self.keys()) keys = sorted(self.keys())
s = torch.tensor([float(self[k]) for k in keys], s = torch.tensor([float(self[k]) for k in keys], device=device)
device=device)
dist.all_reduce(s, op=dist.ReduceOp.SUM) dist.all_reduce(s, op=dist.ReduceOp.SUM)
for k, v in zip(keys, s.cpu().tolist()): for k, v in zip(keys, s.cpu().tolist()):
self[k] = v self[k] = v
def write_summary( def write_summary(
self, self, tb_writer: SummaryWriter, prefix: str, batch_idx: int,
tb_writer: SummaryWriter,
prefix: str,
batch_idx: int,
) -> None: ) -> None:
"""Add logging information to a TensorBoard writer. """Add logging information to a TensorBoard writer.