-
Notifications
You must be signed in to change notification settings - Fork 48
/
Copy pathlightning.py
executable file
·158 lines (131 loc) · 6.96 KB
/
lightning.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
import torch
import torchaudio
from cosine import WarmupCosineScheduler
from datamodule.transforms import TextTransform
from espnet.nets.batch_beam_search import BatchBeamSearch
from espnet.nets.pytorch_backend.e2e_asr_conformer import E2E
from espnet.nets.scorers.length_bonus import LengthBonus
from pytorch_lightning import LightningModule
def compute_word_level_distance(seq1, seq2):
seq1, seq2 = seq1.lower().split(), seq2.lower().split()
return torchaudio.functional.edit_distance(seq1, seq2)
class ModelModule(LightningModule):
def __init__(self, args):
super().__init__()
self.args = args
self.save_hyperparameters(args)
self.modality = args.modality
self.text_transform = TextTransform()
self.token_list = self.text_transform.token_list
self.model = E2E(len(self.token_list), self.modality, ctc_weight=getattr(args, "ctc_weight", 0.1))
# -- initialise
if getattr(args, "pretrained_model_path", None):
ckpt = torch.load(args.pretrained_model_path, map_location=lambda storage, loc: storage)
if getattr(args, "transfer_frontend", False):
tmp_ckpt = {k: v for k, v in ckpt["model_state_dict"].items() if k.startswith("trunk.") or k.startswith("frontend3D.")}
self.model.frontend.load_state_dict(tmp_ckpt)
print("Pretrained weights of the frontend component are loaded successfully.")
elif getattr(args, "transfer_encoder", False):
tmp_ckpt = {k.replace("frontend.",""):v for k,v in ckpt.items() if k.startswith("frontend.")}
self.model.frontend.load_state_dict(tmp_ckpt)
tmp_ckpt = {k.replace("proj_encoder.",""):v for k,v in ckpt.items() if k.startswith("proj_encoder.")}
self.model.proj_encoder.load_state_dict(tmp_ckpt)
tmp_ckpt = {k.replace("encoder.",""):v for k,v in ckpt.items() if k.startswith("encoder.")}
self.model.encoder.load_state_dict(tmp_ckpt)
print("Pretrained weights of the frontend, proj_encoder and encoder component are loaded successfully.")
else:
self.model.load_state_dict(ckpt)
print("Pretrained weights of the full model are loaded successfully.")
def configure_optimizers(self):
optimizer = torch.optim.AdamW(self.model.parameters(), lr=self.args.lr, weight_decay=self.args.weight_decay, betas=(0.9, 0.98))
scheduler = WarmupCosineScheduler(optimizer, self.args.warmup_epochs, self.args.max_epochs, len(self.trainer.datamodule.train_dataloader()) / self.trainer.num_devices / self.trainer.num_nodes)
scheduler = {"scheduler": scheduler, "interval": "step"}
return [optimizer], [scheduler]
def forward(self, sample):
self.beam_search = get_beam_search_decoder(self.model, self.token_list)
x = self.model.frontend(sample.unsqueeze(0))
x = self.model.proj_encoder(x)
enc_feat, _ = self.model.encoder(x, None)
enc_feat = enc_feat.squeeze(0)
nbest_hyps = self.beam_search(enc_feat)
nbest_hyps = [h.asdict() for h in nbest_hyps[: min(len(nbest_hyps), 1)]]
predicted_token_id = torch.tensor(list(map(int, nbest_hyps[0]["yseq"][1:])))
predicted = self.text_transform.post_process(predicted_token_id).replace("<eos>", "")
return predicted
def validation_step(self, batch, batch_idx):
return self._step(batch, batch_idx, step_type="val")
def test_step(self, sample, sample_idx):
x = self.model.frontend(sample["input"].unsqueeze(0))
x = self.model.proj_encoder(x)
enc_feat, _ = self.model.encoder(x, None)
enc_feat = enc_feat.squeeze(0)
nbest_hyps = self.beam_search(enc_feat)
nbest_hyps = [h.asdict() for h in nbest_hyps[: min(len(nbest_hyps), 1)]]
predicted_token_id = torch.tensor(list(map(int, nbest_hyps[0]["yseq"][1:])))
predicted = self.text_transform.post_process(predicted_token_id).replace("<eos>", "")
actual_token_id = sample["target"]
actual = self.text_transform.post_process(actual_token_id)
self.total_edit_distance += compute_word_level_distance(actual, predicted)
self.total_length += len(actual.split())
return
def training_step(self, batch, batch_idx):
loss = self._step(batch, batch_idx, "train")
batch_size = batch["inputs"].size(0)
batch_sizes = self.all_gather(batch_size)
loss *= batch_sizes.size(0) / batch_sizes.sum() # world size / batch size
self.log("monitoring_step", torch.tensor(self.global_step, dtype=torch.float32))
return loss
def _step(self, batch, batch_idx, step_type):
loss, loss_ctc, loss_att, acc = self.model(batch["inputs"], batch["input_lengths"], batch["targets"])
batch_size = len(batch["inputs"])
if step_type == "train":
self.log("loss", loss, on_step=True, on_epoch=True, batch_size=batch_size)
self.log("loss_ctc", loss_ctc, on_step=False, on_epoch=True, batch_size=batch_size, sync_dist=True)
self.log("loss_att", loss_att, on_step=False, on_epoch=True, batch_size=batch_size, sync_dist=True)
self.log("decoder_acc", acc, on_step=True, on_epoch=True, batch_size=batch_size, sync_dist=True)
else:
self.log("loss_val", loss, batch_size=batch_size, sync_dist=True)
self.log("loss_ctc_val", loss_ctc, batch_size=batch_size, sync_dist=True)
self.log("loss_att_val", loss_att, batch_size=batch_size, sync_dist=True)
self.log("decoder_acc_val", acc, batch_size=batch_size, sync_dist=True)
if step_type == "train":
self.log("monitoring_step", torch.tensor(self.global_step, dtype=torch.float32))
return loss
def on_test_epoch_start(self):
self.total_length = 0
self.total_edit_distance = 0
self.text_transform = TextTransform()
self.beam_search = get_beam_search_decoder(self.model, self.token_list)
def on_test_epoch_end(self):
self.log("wer", self.total_edit_distance / self.total_length)
def get_beam_search_decoder(
model,
token_list,
rnnlm=None,
rnnlm_conf=None,
penalty=0,
ctc_weight=0.1,
lm_weight=0.0,
beam_size=40,
):
sos = model.odim - 1
eos = model.odim - 1
scorers = model.scorers()
scorers["lm"] = None
scorers["length_bonus"] = LengthBonus(len(token_list))
weights = {
"decoder": 1.0 - ctc_weight,
"ctc": ctc_weight,
"lm": lm_weight,
"length_bonus": penalty,
}
return BatchBeamSearch(
beam_size=beam_size,
vocab_size=len(token_list),
weights=weights,
scorers=scorers,
sos=sos,
eos=eos,
token_list=token_list,
pre_beam_score_key=None if ctc_weight == 1.0 else "decoder",
)