-
Notifications
You must be signed in to change notification settings - Fork 0
/
trainer_seq2seq.py
315 lines (270 loc) · 12.6 KB
/
trainer_seq2seq.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
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
# Copyright 2020 The HuggingFace Team. All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import os
from typing import Any, Dict, List, Optional, Tuple, Union
import torch
from torch import nn
from torch.utils.data import Dataset
from trainer import Trainer
from transformers.utils import logging
from transformers.training_args import TrainingArguments
from transformers.deepspeed import is_deepspeed_zero3_enabled
from transformers.trainer_utils import PredictionOutput, PREFIX_CHECKPOINT_DIR
from transformers.trainer_callback import TrainerCallback, TrainerControl, TrainerState
logger = logging.get_logger(__name__)
class SavePeftModelCallback(TrainerCallback):
def on_save(
self,
args: TrainingArguments,
state: TrainerState,
control: TrainerControl,
**kwargs,
):
checkpoint_folder = os.path.join(
args.output_dir, f"{PREFIX_CHECKPOINT_DIR}-{state.global_step}"
)
peft_model_path = os.path.join(checkpoint_folder, "adapter_model")
kwargs["model"].save_pretrained(peft_model_path)
pytorch_model_path = os.path.join(checkpoint_folder, "pytorch_model.bin")
if os.path.exists(pytorch_model_path):
os.remove(pytorch_model_path)
return control
class Seq2SeqTrainer(Trainer):
def evaluate(
self,
eval_dataset: Optional[Dataset] = None,
ignore_keys: Optional[List[str]] = None,
metric_key_prefix: str = "eval",
**gen_kwargs,
) -> Dict[str, float]:
"""
Run evaluation and returns metrics.
The calling script will be responsible for providing a method to compute metrics, as they are task-dependent
(pass it to the init `compute_metrics` argument).
You can also subclass and override this method to inject custom behavior.
Args:
eval_dataset (`Dataset`, *optional*):
Pass a dataset if you wish to override `self.eval_dataset`. If it is an [`~datasets.Dataset`], columns
not accepted by the `model.forward()` method are automatically removed. It must implement the `__len__`
method.
ignore_keys (`List[str]`, *optional*):
A list of keys in the output of your model (if it is a dictionary) that should be ignored when
gathering predictions.
metric_key_prefix (`str`, *optional*, defaults to `"eval"`):
An optional prefix to be used as the metrics key prefix. For example the metrics "bleu" will be named
"eval_bleu" if the prefix is `"eval"` (default)
max_length (`int`, *optional*):
The maximum target length to use when predicting with the generate method.
num_beams (`int`, *optional*):
Number of beams for beam search that will be used when predicting with the generate method. 1 means no
beam search.
gen_kwargs:
Additional `generate` specific kwargs.
Returns:
A dictionary containing the evaluation loss and the potential metrics computed from the predictions. The
dictionary also contains the epoch number which comes from the training state.
"""
gen_kwargs = gen_kwargs.copy()
if (
gen_kwargs.get("max_length") is None
and gen_kwargs.get("max_new_tokens") is None
):
gen_kwargs["max_length"] = self.args.generation_max_length
gen_kwargs["num_beams"] = (
gen_kwargs["num_beams"]
if gen_kwargs.get("num_beams") is not None
else self.args.generation_num_beams
)
self._gen_kwargs = gen_kwargs
return super().evaluate(
eval_dataset, ignore_keys=ignore_keys, metric_key_prefix=metric_key_prefix
)
def predict(
self,
test_dataset: Dataset,
ignore_keys: Optional[List[str]] = None,
metric_key_prefix: str = "test",
**gen_kwargs,
) -> PredictionOutput:
"""
Run prediction and returns predictions and potential metrics.
Depending on the dataset and your use case, your test dataset may contain labels. In that case, this method
will also return metrics, like in `evaluate()`.
Args:
test_dataset (`Dataset`):
Dataset to run the predictions on. If it is a [`~datasets.Dataset`], columns not accepted by the
`model.forward()` method are automatically removed. Has to implement the method `__len__`
ignore_keys (`List[str]`, *optional*):
A list of keys in the output of your model (if it is a dictionary) that should be ignored when
gathering predictions.
metric_key_prefix (`str`, *optional*, defaults to `"eval"`):
An optional prefix to be used as the metrics key prefix. For example the metrics "bleu" will be named
"eval_bleu" if the prefix is `"eval"` (default)
max_length (`int`, *optional*):
The maximum target length to use when predicting with the generate method.
num_beams (`int`, *optional*):
Number of beams for beam search that will be used when predicting with the generate method. 1 means no
beam search.
gen_kwargs:
Additional `generate` specific kwargs.
<Tip>
If your predictions or labels have different sequence lengths (for instance because you're doing dynamic
padding in a token classification task) the predictions will be padded (on the right) to allow for
concatenation into one array. The padding index is -100.
</Tip>
Returns: *NamedTuple* A namedtuple with the following keys:
- predictions (`np.ndarray`): The predictions on `test_dataset`.
- label_ids (`np.ndarray`, *optional*): The labels (if the dataset contained some).
- metrics (`Dict[str, float]`, *optional*): The potential dictionary of metrics (if the dataset contained
labels).
"""
gen_kwargs = gen_kwargs.copy()
if (
gen_kwargs.get("max_length") is None
and gen_kwargs.get("max_new_tokens") is None
):
gen_kwargs["max_length"] = self.args.generation_max_length
gen_kwargs["num_beams"] = (
gen_kwargs["num_beams"]
if gen_kwargs.get("num_beams") is not None
else self.args.generation_num_beams
)
self._gen_kwargs = gen_kwargs
return super().predict(
test_dataset, ignore_keys=ignore_keys, metric_key_prefix=metric_key_prefix
)
def prediction_step(
self,
model: nn.Module,
inputs: Dict[str, Union[torch.Tensor, Any]],
prediction_loss_only: bool,
ignore_keys: Optional[List[str]] = None,
) -> Tuple[Optional[float], Optional[torch.Tensor], Optional[torch.Tensor]]:
"""
Perform an evaluation step on `model` using `inputs`.
Subclass and override to inject custom behavior.
Args:
model (`nn.Module`):
The model to evaluate.
inputs (`Dict[str, Union[torch.Tensor, Any]]`):
The inputs and targets of the model.
The dictionary will be unpacked before being fed to the model. Most models expect the targets under the
argument `labels`. Check your model's documentation for all accepted arguments.
prediction_loss_only (`bool`):
Whether or not to return the loss only.
Return:
Tuple[Optional[float], Optional[torch.Tensor], Optional[torch.Tensor]]: A tuple with the loss, logits and
labels (each being optional).
"""
if not self.args.predict_with_generate or prediction_loss_only:
return super().prediction_step(
model,
inputs,
prediction_loss_only=prediction_loss_only,
ignore_keys=ignore_keys,
)
has_labels = "labels" in inputs
inputs = self._prepare_inputs(inputs)
# XXX: adapt synced_gpus for fairscale as well
gen_kwargs = self._gen_kwargs.copy()
if (
gen_kwargs.get("max_length") is None
and gen_kwargs.get("max_new_tokens") is None
):
gen_kwargs["max_length"] = self.model.config.max_length
gen_kwargs["num_beams"] = (
gen_kwargs["num_beams"]
if gen_kwargs.get("num_beams") is not None
else self.model.config.num_beams
)
default_synced_gpus = True if is_deepspeed_zero3_enabled() else False
gen_kwargs["synced_gpus"] = (
gen_kwargs["synced_gpus"]
if gen_kwargs.get("synced_gpus") is not None
else default_synced_gpus
)
if "attention_mask" in inputs:
gen_kwargs["attention_mask"] = inputs.get("attention_mask", None)
if "position_ids" in inputs:
gen_kwargs["position_ids"] = inputs.get("position_ids", None)
if "global_attention_mask" in inputs:
gen_kwargs["global_attention_mask"] = inputs.get(
"global_attention_mask", None
)
# prepare generation inputs
# some encoder-decoder models can have varying encoder's and thus
# varying model input names
if (
hasattr(self.model, "encoder")
and self.model.encoder.main_input_name != self.model.main_input_name
):
generation_inputs = inputs[self.model.encoder.main_input_name]
else:
generation_inputs = inputs[self.model.main_input_name]
gen_kwargs["input_ids"] = generation_inputs
generated_tokens = self.model.generate(**gen_kwargs)
generated_tokens = generated_tokens[:, generation_inputs.size()[-1] :]
# in case the batch is shorter than max length, the output should be padded
if (
gen_kwargs.get("max_length") is not None
and generated_tokens.shape[-1] < gen_kwargs["max_length"]
):
generated_tokens = self._pad_tensors_to_max_len(
generated_tokens, gen_kwargs["max_length"]
)
elif gen_kwargs.get("max_new_tokens") is not None and generated_tokens.shape[
-1
] < (gen_kwargs["max_new_tokens"] + 1):
generated_tokens = self._pad_tensors_to_max_len(
generated_tokens, gen_kwargs["max_new_tokens"] + 1
)
loss = None
if self.args.prediction_loss_only:
return (loss, None, None)
if has_labels:
labels = inputs["labels"]
if (
gen_kwargs.get("max_length") is not None
and labels.shape[-1] < gen_kwargs["max_length"]
):
labels = self._pad_tensors_to_max_len(labels, gen_kwargs["max_length"])
elif gen_kwargs.get("max_new_tokens") is not None and labels.shape[-1] < (
gen_kwargs["max_new_tokens"] + 1
):
labels = self._pad_tensors_to_max_len(
labels, (gen_kwargs["max_new_tokens"] + 1)
)
else:
labels = None
return (loss, generated_tokens, labels)
def _pad_tensors_to_max_len(self, tensor, max_length):
if self.tokenizer is not None and hasattr(self.tokenizer, "pad_token_id"):
# If PAD token is not defined at least EOS token has to be defined
pad_token_id = (
self.tokenizer.pad_token_id
if self.tokenizer.pad_token_id is not None
else self.tokenizer.eos_token_id
)
else:
if self.model.config.pad_token_id is not None:
pad_token_id = self.model.config.pad_token_id
else:
raise ValueError(
"Pad_token_id must be set in the configuration of the model, in order to pad tensors"
)
padded_tensor = pad_token_id * torch.ones(
(tensor.shape[0], max_length), dtype=tensor.dtype, device=tensor.device
)
padded_tensor[:, : tensor.shape[-1]] = tensor
return padded_tensor