-
Notifications
You must be signed in to change notification settings - Fork 0
/
train.py
139 lines (120 loc) · 4.37 KB
/
train.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
import multiprocessing
import numpy as np
import torch
from datasets import load_dataset
from transformers import (
AutoModelForMaskedLM,
AutoTokenizer,
DataCollatorForLanguageModeling,
EarlyStoppingCallback,
TrainingArguments,
Trainer,
)
from utils import (
seed_everything,
load_config,
tokenize_function,
group_texts,
whole_word_masking_data_collator,
)
CFG = load_config()
seed_everything(CFG.seed)
tokenizer = AutoTokenizer.from_pretrained(CFG.model_checkpoint)
model = AutoModelForMaskedLM.from_pretrained(CFG.model_checkpoint)
dset = load_dataset(CFG.data_path)
# Use batched=True to activate fast multithreading!
tokenized_datasets = dset.map(
tokenize_function,
batched=True,
remove_columns=["text"],
num_proc=multiprocessing.cpu_count() // 2,
)
# Slicing produces a list of lists for each feature
tokenized_samples = tokenized_datasets["train"][:3]
for idx, sample in enumerate(tokenized_samples["input_ids"]):
print(f"'>>> email {idx} length: {len(sample)}'")
concatenated_examples = {k: sum(tokenized_samples[k], []) for k in tokenized_samples.keys()}
total_length = len(concatenated_examples["input_ids"])
print(f"'>>> Concatenated email length: {total_length}'")
chunks = {
k: [t[i : i + CFG.chunk_size] for i in range(0, total_length, CFG.chunk_size)]
for k, t in concatenated_examples.items()
}
for chunk in chunks["input_ids"]:
print(f"'>>> Chunk length: {len(chunk)}'")
lm_datasets = tokenized_datasets.map(group_texts, batched=True, num_proc=6)
print(tokenizer.decode(lm_datasets["train"][1]["input_ids"]))
print(tokenizer.decode(lm_datasets["train"][1]["labels"]))
data_collator = DataCollatorForLanguageModeling(
tokenizer=tokenizer, mlm_probability=CFG.masking_probability
)
samples = [lm_datasets["train"][i] for i in range(2)]
for sample in samples:
_ = sample.pop("word_ids")
for chunk in data_collator(samples)["input_ids"]:
print(f"\n'>>> {tokenizer.decode(chunk)}'")
samples = [lm_datasets["train"][i] for i in range(2)]
batch = whole_word_masking_data_collator(samples)
for chunk in batch["input_ids"]:
print(f"\n'>>> {tokenizer.decode(chunk)}'")
downsampled_dataset = lm_datasets["train"].train_test_split(
train_size=CFG.train_size, test_size=CFG.test_size, seed=CFG.seed
)
print(downsampled_dataset)
# Show the training loss with every epoch
epoch_step = len(downsampled_dataset["train"]) // (CFG.batch_size * torch.cuda.device_count())
model_name = CFG.model_checkpoint.split("/")[-1]
if CFG.early_stop <= -1:
# no early stopping
training_args = TrainingArguments(
output_dir=f"{model_name}-email-finetuned",
overwrite_output_dir=True,
learning_rate=CFG.learning_rate,
weight_decay=CFG.weight_decay,
warmup_ratio=CFG.warm_up_ratio,
per_device_train_batch_size=CFG.batch_size,
per_device_eval_batch_size=CFG.batch_size,
evaluation_strategy="epoch",
save_strategy="epoch",
num_train_epochs=CFG.num_train_epochs,
do_eval=True,
push_to_hub=True,
fp16=CFG.fp16,
gradient_checkpointing=CFG.gradient_checkpointing,
logging_steps=CFG.logging_steps,
save_total_limit=CFG.save_total_limit,
metric_for_best_model="loss",
load_best_model_at_end=True,
)
elif CFG.early_stop > 0:
training_args = TrainingArguments(
output_dir=f"{model_name}-email-finetuned",
overwrite_output_dir=True,
learning_rate=CFG.learning_rate,
weight_decay=CFG.weight_decay,
warmup_ratio=CFG.warm_up_ratio,
per_device_train_batch_size=CFG.batch_size,
per_device_eval_batch_size=CFG.batch_size,
evaluation_strategy="steps",
eval_steps=epoch_step,
save_steps=epoch_step,
num_train_epochs=CFG.num_train_epochs,
do_eval=True,
push_to_hub=True,
fp16=CFG.fp16,
gradient_checkpointing=CFG.gradient_checkpointing,
logging_steps=CFG.logging_steps,
save_total_limit=CFG.save_total_limit,
metric_for_best_model="loss",
load_best_model_at_end=True,
)
trainer = Trainer(
model=model,
args=training_args,
train_dataset=downsampled_dataset["train"],
eval_dataset=downsampled_dataset["test"],
data_collator=data_collator,
callbacks=[EarlyStoppingCallback(early_stopping_patience=CFG.early_stop)],
)
trainer.train()
trainer.push_to_hub()