-
Notifications
You must be signed in to change notification settings - Fork 1
/
platonic_opt.py
175 lines (142 loc) · 5.03 KB
/
platonic_opt.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
import argparse
import os
import sys
import torch
import transformers
from datasets import load_dataset
from peft import LoraConfig, TaskType, get_peft_model
from transformers import (
OPTForCausalLM,
AutoTokenizer,
Trainer,
TrainingArguments,
)
def get_device():
if torch.cuda.is_available():
return "cuda"
elif torch.backends.mps.is_available():
return "mps"
else:
return "cpu"
def load_plato_dataset(
tokenizer: AutoTokenizer,
data_dir: str,
max_sequence_length: int = 2048,
seed: int = 42,
):
def _tokenize_document(e):
outputs = tokenizer(
e["text"],
truncation=True,
padding="max_length",
max_length=max_sequence_length,
return_length=True,
return_overflowing_tokens=True,
)
input_batch = []
attention_mask_batch = []
for length, input_ids, attention_mask in zip(
outputs["length"], outputs["input_ids"], outputs["attention_mask"]
):
if length == max_sequence_length:
input_batch.append(input_ids)
attention_mask_batch.append(attention_mask)
return {
"input_ids": input_batch,
"attention_mask": attention_mask_batch,
"labels": input_batch,
}
# use "Apology" as validation text
VALIDATION_TEXT_ID = "1656.txt"
train_data_files = [x for x in os.listdir(data_dir) if x != VALIDATION_TEXT_ID]
dataset = load_dataset(
"text",
keep_linebreaks=False,
sample_by="document",
data_files={
"train": [os.path.join(data_dir, f) for f in train_data_files],
"validation": os.path.join(data_dir, VALIDATION_TEXT_ID),
},
)
dataset = dataset.map(
_tokenize_document, batched=True, remove_columns=["text"]
).shuffle(seed=seed)
dataset = dataset.with_format("torch")
return dataset["train"], dataset["validation"]
def main(args):
parser = argparse.ArgumentParser()
parser.add_argument("--model-name", type=str, required=True)
parser.add_argument("--output-dir", type=str, required=True)
parser.add_argument("--data-dir", type=str, required=True)
parser.add_argument("--epochs", type=int, default=2)
parser.add_argument("--batch-size", type=int, default=8)
parser.add_argument("--lr", type=float, default=5e-5)
parser.add_argument("--weight-decay", type=float, default=0.01)
parser.add_argument("--warmup", type=float, default=0.1)
parser.add_argument("--r", type=int, default=4)
parser.add_argument(
"--fine-tuning",
type=str,
choices=["full", "top", "top2", "lora"],
default="full",
)
args = parser.parse_args(args)
os.makedirs(args.output_dir, exist_ok=True)
print(f"Loading model {args.model_name}...")
model = OPTForCausalLM.from_pretrained(args.model_name)
# freeze all but the final encoder
if args.fine_tuning == "top":
for param in model.model.parameters():
param.requires_grad = False
for param in model.model.decoder.layers[-1].parameters():
param.requires_grad = True
# freeze all but the last two encoders
elif args.fine_tuning == "top2":
for param in model.model.parameters():
param.requires_grad = False
for param in model.model.decoder.layers[-2:].parameters():
param.requires_grad = True
elif args.fine_tuning == "lora":
lora_config = LoraConfig(
task_type=TaskType.CAUSAL_LM,
inference_mode=False,
r=args.r,
lora_alpha=args.r,
lora_dropout=0.1,
)
model = get_peft_model(model, lora_config)
model.print_trainable_parameters()
tokenizer = AutoTokenizer.from_pretrained(args.model_name)
print("Loading Plato's texts...")
train_dataset, eval_dataset = load_plato_dataset(tokenizer, args.data_dir)
training_args = TrainingArguments(
output_dir=args.output_dir,
num_train_epochs=args.epochs,
per_device_train_batch_size=args.batch_size,
per_device_eval_batch_size=args.batch_size,
weight_decay=args.weight_decay,
learning_rate=args.lr,
warmup_ratio=args.warmup,
save_steps=1_000,
eval_steps=1_000,
save_total_limit=5,
use_mps_device=get_device() == "mps",
)
trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_dataset,
eval_dataset=eval_dataset,
)
initial_metrics = trainer.evaluate()
trainer.save_metrics("initial", initial_metrics)
print("Starting to train...")
transformers.logging.set_verbosity_info()
trainer.train()
final_checkpoint = os.path.join(args.output_dir, "final-model".format(args.epochs))
os.makedirs(final_checkpoint, exist_ok=True)
trainer.save_model(final_checkpoint)
final_metrics = trainer.evaluate()
trainer.save_metrics("eval", final_metrics)
if __name__ == "__main__":
main(sys.argv[1:])