-
Notifications
You must be signed in to change notification settings - Fork 28.5k
/
Copy pathdistillation.py
executable file
·296 lines (252 loc) · 13.3 KB
/
distillation.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
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
#!/usr/bin/env python
import argparse
import gc
import os
import sys
from pathlib import Path
from typing import List
import pytorch_lightning as pl
import torch
from torch import nn
from torch.nn import functional as F
from finetune import SummarizationModule, TranslationModule
from finetune import main as ft_main
from make_student import create_student_by_copying_alternating_layers, get_layers_to_supervise
from transformers import AutoModelForSeq2SeqLM, MBartTokenizer, T5ForConditionalGeneration
from transformers.modeling_bart import shift_tokens_right
from utils import calculate_bleu, freeze_params, label_smoothed_nll_loss, pickle_load, use_task_specific_params
# need the parent dir module
sys.path.insert(2, str(Path(__file__).resolve().parents[1]))
from lightning_base import generic_train # noqa
class BartSummarizationDistiller(SummarizationModule):
"""Supports Bart, Pegasus and other models that inherit from Bart."""
loss_names = ["loss", "ce_loss", "mlm_loss", "enc_mse_loss", "hid_loss_enc", "hid_loss_dec"]
def __init__(self, hparams):
assert Path(hparams.data_dir).exists()
self.output_dir = Path(hparams.output_dir)
self.output_dir.mkdir(exist_ok=True)
save_dir = self.output_dir.joinpath("student")
hparams.model_name_or_path = str(save_dir) # Tell lightning we are training the student
teacher = AutoModelForSeq2SeqLM.from_pretrained(hparams.teacher).eval()
use_task_specific_params(teacher, hparams.task) # We copy good generation parameters to student by default
student, e_layer_ids, d_layer_ids = create_student_by_copying_alternating_layers(
teacher, e=hparams.student_encoder_layers, d=hparams.student_decoder_layers, save_path=save_dir
)
if hparams.length_penalty != -1:
student.config.length_penalty = hparams.length_penalty
super().__init__(hparams, model=student, config=student.config)
self.e_layer_ids, self.d_layer_ids = e_layer_ids, d_layer_ids # type: List[int], List[int]
self.different_encoder = hparams.student_encoder_layers != teacher.config.encoder_layers
self.different_decoder = hparams.student_decoder_layers != teacher.config.decoder_layers
self.teacher = teacher
freeze_params(self.teacher)
if not self.different_encoder: # To save RAM, delete teacher encoder and freeze student encoder.
try:
del self.teacher.model.encoder
except AttributeError: # T5
del self.teacher.encoder
# Intermediate supervision: Decide which layers to supervise
if hparams.supervise_forward:
self.d_matches = get_layers_to_supervise(
n_student=len(self.d_layer_ids), n_teacher=self.teacher.config.decoder_layers
)
else:
self.d_matches = self.d_layer_ids
self.ce_loss_fct = nn.KLDivLoss(reduction="batchmean")
self.temperature = 2.0
self.alpha_mlm = hparams.alpha_mlm
self.alpha_ce = hparams.alpha_ce
self.alpha_hid = hparams.alpha_hid
self.alpha_encoder_loss = hparams.alpha_encoder_loss
gc.collect()
torch.cuda.empty_cache()
def calc_mse_loss(self, teacher_outputs: torch.Tensor, student_outputs: torch.Tensor, mask) -> torch.FloatTensor:
"""Supervise MSE(teacher.encoder_outputs, student.encoder_outputs)."""
# raise NotImplementedError()
if mask is not None:
# mask has False at padding_idx
sel_mask = mask[:, :, None].expand_as(student_outputs).bool()
s_logits_slct = torch.masked_select(student_outputs, sel_mask)
t_logits_slct = torch.masked_select(teacher_outputs, sel_mask)
else:
t_logits_slct = teacher_outputs
s_logits_slct = student_outputs
return F.mse_loss(s_logits_slct, t_logits_slct)
def calc_ce_loss(self, mask, s_logits, t_logits):
"""Copy pasted from distillbert (transformers/examples/distillation/)"""
# mask has False at padding_idx
sel_mask = mask[:, :, None].expand_as(s_logits)
vocab_size = s_logits.size(-1)
s_logits_slct = torch.masked_select(s_logits, sel_mask) # (bs * seq_length * voc_size) modulo the 1s in mask
t_logits_slct = torch.masked_select(t_logits, sel_mask) # (bs * seq_length * voc_size) modulo the 1s in mask
s_logits_slct = s_logits_slct.view(-1, vocab_size) # (bs * seq_length, voc_size) modulo the 1s in mask
t_logits_slct = t_logits_slct.view(-1, vocab_size) # (bs * seq_length, voc_size) modulo the 1s in mask
assert t_logits_slct.size() == s_logits_slct.size()
loss_ce = (
self.ce_loss_fct(
F.log_softmax(s_logits_slct / self.temperature, dim=-1),
F.softmax(t_logits_slct / self.temperature, dim=-1),
)
* (self.temperature) ** 2
)
return loss_ce
@staticmethod
def add_model_specific_args(parser, root_dir):
SummarizationModule.add_model_specific_args(parser, root_dir)
add_distill_args(parser)
return parser
def _step(self, batch):
# assert is_frozen(self.teacher) copied_decoder_layers
pad_token_id = self.tokenizer.pad_token_id
input_ids, src_mask, labels = batch["input_ids"], batch["attention_mask"], batch["labels"]
if isinstance(self.model, T5ForConditionalGeneration):
decoder_input_ids = self.model._shift_right(labels)
else:
decoder_input_ids = shift_tokens_right(labels, pad_token_id)
# noinspection PyCallingNonCallable
lm_logits, dec_hidden, enc_outputs, enc_hidden_state = self(
input_ids,
attention_mask=src_mask,
decoder_input_ids=decoder_input_ids,
output_hidden_states=True,
output_attentions=False,
use_cache=False,
) # TODO(@sshleifer): return_dict=True cleanup
# Same cross entropy vs. label smoothing logic as finetune.py
assert lm_logits.shape[-1] == self.model.config.vocab_size
if self.hparams.label_smoothing == 0:
# Same behavior as modeling_bart.py, besides ignoring pad_token_id
loss_fct = torch.nn.CrossEntropyLoss(ignore_index=pad_token_id)
student_lm_loss = loss_fct(lm_logits.view(-1, lm_logits.shape[-1]), labels.view(-1))
else:
lprobs = torch.nn.functional.log_softmax(lm_logits, dim=-1)
student_lm_loss, _ = label_smoothed_nll_loss(
lprobs, labels, self.hparams.label_smoothing, ignore_index=pad_token_id
)
def zero_tensor():
return torch.tensor(0.0).type_as(student_lm_loss)
loss_encoder, hid_loss_enc, hid_loss_dec = zero_tensor(), zero_tensor(), zero_tensor()
if self.different_encoder:
with torch.no_grad():
teacher_enc_outputs, teacher_enc_hid, _ = self.teacher.get_encoder()(
input_ids, attention_mask=src_mask, output_hidden_states=True
)
# DEPRECATE THIS
if self.hparams.alpha_encoder_loss > 0:
loss_encoder = self.calc_mse_loss(enc_outputs, teacher_enc_outputs, src_mask)
hid_loss_enc = self.calc_hidden_loss(src_mask, enc_hidden_state, teacher_enc_hid, self.e_layer_ids)
teacher_enc_outputs = (enc_outputs,)
assert isinstance(teacher_enc_outputs, tuple), type(teacher_enc_outputs)
with torch.no_grad():
tloss, tlogits, tdec_hidden, _ = self.teacher(
input_ids,
attention_mask=src_mask,
encoder_outputs=teacher_enc_outputs,
decoder_input_ids=decoder_input_ids,
lm_labels=labels,
output_hidden_states=True,
)
dec_mask = decoder_input_ids.ne(pad_token_id)
loss_ce = self.calc_ce_loss(dec_mask, lm_logits, tlogits)
if self.alpha_hid > 0: # Intermediate supervision of decoder hidden states
hid_loss_dec = self.calc_hidden_loss(
dec_mask, dec_hidden, tdec_hidden, self.d_matches, normalize_hidden=self.hparams.normalize_hidden
)
blended_loss = (
self.alpha_ce * loss_ce
+ self.alpha_mlm * student_lm_loss
+ self.hparams.alpha_encoder_loss * loss_encoder
+ self.hparams.alpha_hid * (hid_loss_enc + hid_loss_dec)
)
return blended_loss, loss_ce, student_lm_loss, loss_encoder, hid_loss_enc, hid_loss_dec
@staticmethod
def calc_hidden_loss(attention_mask, hidden_states, hidden_states_T, matches, normalize_hidden):
"""MSE(student_hid, teacher_hid[matches]). Called "Intermediate supervision" in paper. Inspired by TinyBERT."""
msg = "expected list or tuple for hidden_states, got tensor of shape: "
assert not isinstance(hidden_states, torch.Tensor), f"{msg}{hidden_states.shape}"
assert not isinstance(hidden_states_T, torch.Tensor), f"{msg}{hidden_states_T.shape}"
mask = attention_mask.to(hidden_states[0])
valid_count = mask.sum() * hidden_states[0].size(-1)
student_states = torch.stack([hidden_states[i] for i in range(len(matches))])
teacher_states = torch.stack([hidden_states_T[j] for j in matches])
if normalize_hidden:
student_states = F.layer_norm(student_states, student_states.shape[1:])
teacher_states = F.layer_norm(teacher_states, teacher_states.shape[1:])
mse = F.mse_loss(student_states, teacher_states, reduction="none")
masked_mse = (mse * mask.unsqueeze(0).unsqueeze(-1)).sum() / valid_count
return masked_mse
def add_distill_args(parser):
parser.add_argument("--teacher", type=str)
parser.add_argument("--alpha_ce", default=0.8, type=float)
parser.add_argument("--alpha_mlm", default=0.2, type=float)
parser.add_argument("--alpha_encoder_loss", default=0.0, type=float)
parser.add_argument("--alpha_hid", default=0.0, type=float, required=False)
parser.add_argument("--student_decoder_layers", default=12, type=int, required=False)
parser.add_argument("--student_encoder_layers", default=12, type=int, required=False)
parser.add_argument("--no_teacher", action="store_true", default=False)
parser.add_argument("--length_penalty", type=float, default=-1)
parser.add_argument("--supervise_forward", action="store_true", default=False)
parser.add_argument("--normalize_hidden", action="store_true", default=False)
class BartTranslationDistiller(BartSummarizationDistiller):
"""Supports Mbart, Marian, other models that inherit from Bart."""
mode = "translation"
metric_names = ["bleu"]
default_val_metric = "bleu"
def __init__(self, hparams, **kwargs):
super().__init__(hparams, **kwargs)
assert hparams.src_lang is not None
assert hparams.tgt_lang is not None
self.dataset_kwargs["src_lang"] = hparams.src_lang
self.dataset_kwargs["tgt_lang"] = hparams.tgt_lang
if self.model.config.decoder_start_token_id is None and isinstance(self.tokenizer, MBartTokenizer):
self.decoder_start_token_id = self.tokenizer.lang_code_to_id[hparams.tgt_lang]
def calc_generative_metrics(self, preds, target) -> dict:
return calculate_bleu(preds, target)
@staticmethod
def add_model_specific_args(parser, root_dir):
TranslationModule.add_model_specific_args(parser, root_dir)
add_distill_args(parser)
return parser
def create_module(args):
if args.no_teacher:
module_cls = TranslationModule if "translation" in args.task else SummarizationModule
else: # DISTILL WITH TEACHER
module_cls = BartTranslationDistiller if "translation" in args.task else BartSummarizationDistiller
args.setup_cls: str = module_cls.__name__
print(f"using module {args.setup_cls}")
model = module_cls(args)
return model
def evaluate_checkpoint(ckpt_path: Path, dest_dir=None):
# TODO(SS): DELETE? Better to convert_pl_ckpt_to_hf and run_eval.py
exp_dir = ckpt_path.parent
if dest_dir is None:
dest_dir = exp_dir
clash = list(dest_dir.glob("test_generations*"))
if clash:
print(f"SKIPPING to avoid overwriting {clash}")
ckpt = torch.load(ckpt_path, map_location="cpu")
if "hparams" in ckpt:
args = argparse.Namespace(**ckpt["hparams"])
else:
args = argparse.Namespace(**pickle_load(exp_dir / "hparams.pkl"))
args.resume_from_checkpoint = str(ckpt_path)
args.do_train = False
args.output_dir = str(dest_dir)
args.n_gpu = 1
args.eval_batch_size = 16
Path(args.output_dir).mkdir(exist_ok=True)
model = create_module(args)
trainer: pl.Trainer = generic_train(model, args, early_stopping_callback=False)
trainer.test(model)
def distill_main(args):
Path(args.output_dir).mkdir(exist_ok=True)
if len(os.listdir(args.output_dir)) > 3 and args.do_train:
raise ValueError("Output directory ({}) already exists and is not empty.".format(args.output_dir))
model = create_module(args)
return ft_main(args, model=model)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser = pl.Trainer.add_argparse_args(parser)
parser = BartSummarizationDistiller.add_model_specific_args(parser, os.getcwd())
args = parser.parse_args()
distill_main(args)