-
Notifications
You must be signed in to change notification settings - Fork 1
/
preprocess_cot.py
181 lines (142 loc) · 6.75 KB
/
preprocess_cot.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
import numpy as np
import os
import json
import argparse
from datasets import load_dataset, concatenate_datasets, Dataset
import transformers
from transformers import AutoTokenizer
from verify_cot import Verifier
from tqdm import tqdm
from termcolor import cprint
transformers.utils.move_cache()
parser = argparse.ArgumentParser()
parser.add_argument('--dataset', type=str, default="medqa_usmle_hf",
choices=["medqa_usmle_hf", "strategyqa", "obqa"])
args = parser.parse_args()
### Experiment config
model_id = "google/flan-t5-large" # Hugging Face Model Id
text_column = "question" # column of input text is
summary_column = "reasoning" # column of the output text
if args.dataset == "medqa_usmle_hf":
prompt_template = f"The following are multiple-choice questions about medical knowledge. Generate a step-by-step explanations for each question:\n{{input}}\nExplanation:\n"
data_template = "{} A. {} B. {} C. {} D. {}"
elif args.dataset == "strategyqa":
prompt_template = f"The following are multiple-choice questions. Generate a step-by-step explanations for each question:\n{{input}}\nExplanation:\n"
data_template = "{} A. {} B. {}"
elif args.dataset == "obqa":
prompt_template = f"The following are multiple-choice questions about commonsense knowledge. Generate a step-by-step explanations for each question:\n{{input}}\nExplanation:\n"
data_template = "{} A. {} B. {} C. {} D. {}"
else:
raise NotImplementedError
def format_question(args, data):
if args.dataset in ["medqa_usmle_hf", "obqa"]:
q = data_template.format(data["sent1"], data["ending0"], data["ending1"], data["ending2"], data["ending3"])
elif args.dataset in ["strategyqa"]:
q = data_template.format(data["sent1"], data["ending0"], data["ending1"])
else:
raise NotImplementedError
return q
choices = ["A", "B", "C", "D"]
### Path Definition
if args.dataset == "medqa_usmle_hf":
save_dataset_path = "preprocessed_data/medqa-cot" # local path to save processed dataset
elif args.dataset == "strategyqa":
save_dataset_path = "preprocessed_data/strategyqa-cot"
elif args.dataset == "obqa":
save_dataset_path = "preprocessed_data/obqa-cot"
else:
raise NotImplementedError
os.makedirs(save_dataset_path, exist_ok=True)
cot_dump_path = f"./chain_of_thoughts/{args.dataset}"
os.makedirs(cot_dump_path, exist_ok=True)
fpx = open(os.path.join(cot_dump_path, "train_dump.jsonl"), 'w+')
### CoT Path
cot_dir_path = f"./chatgpt_output/{args.dataset}/train"
print(args.dataset)
# Load raw dataset
dataset_raw = load_dataset("json", data_files=f"data/{args.dataset}/train.json", split="train")
verifier = Verifier(args)
new_dataset = []
cot_dumps = []
for idx, data in enumerate(tqdm(dataset_raw)):
cot_file_path = os.path.join(cot_dir_path, str(idx).zfill(5) + ".json")
if not os.path.exists(cot_file_path):
print(f"Not Exist {cot_file_path}")
print("We expert there is no more cot afterwards in now....")
break
with open(cot_file_path) as f:
cot_data = json.load(f)
correct_cot, wrong_cot = verifier(data, cot_data)
cot_data = {
"correct": correct_cot,
"wrong": wrong_cot,
}
# cot_dumps.append(cot_data)
fpx.write(json.dumps(cot_data))
fpx.write('\n')
for cot in correct_cot:
answer = data["label"]
option = choices[int(answer)]
answer_prompt = f"\nAnswer: {option}"
new_data = {
"question": format_question(args, data),
"reasoning": cot + answer_prompt,
"answer": data["label"]
}
new_dataset.append(new_data)
if len(new_dataset) < 10:
cprint("Question", 'red')
print(new_data["question"])
cprint("Reasoning", 'blue')
print(new_data["reasoning"])
print("+" * 20)
print()
# if len(new_dataset) > 100:
# break
fpx.close()
print(f"New Dataset Size: {len(new_dataset)}")
train_len = int(len(new_dataset) * 0.95)
train_dataset = new_dataset[:train_len]
test_dataset = new_dataset[train_len:] # Split validation set for Reasoning Distillation
with open(os.path.join(save_dataset_path, "train.json"), 'w+') as f:
json.dump(train_dataset, f)
with open(os.path.join(save_dataset_path, "test.json"), 'w+') as f:
json.dump(test_dataset, f)
####### Step 1 done
dataset = load_dataset('json', data_files={'train': os.path.join(save_dataset_path, "train.json"),
'test': os.path.join(save_dataset_path, "test.json")})
tokenizer = AutoTokenizer.from_pretrained(model_id)
print(f"Train dataset size: {len(dataset['train'])}")
print(f"Test dataset size: {len(dataset['test'])}")
prompt_lenght = len(tokenizer(prompt_template.format(input=""))["input_ids"])
max_sample_length = tokenizer.model_max_length - prompt_lenght
print(f"Prompt length: {prompt_lenght}")
print(f"Max input length: {max_sample_length}")
tokenized_inputs = concatenate_datasets([dataset["train"], dataset["test"]]).map(lambda x: tokenizer(x[text_column], truncation=True), batched=True, remove_columns=[text_column, summary_column])
max_source_length = max([len(x) for x in tokenized_inputs["input_ids"]])
max_source_length = min(max_source_length, max_sample_length)
print(f"Max source length: {max_source_length}")
tokenized_targets = concatenate_datasets([dataset["train"], dataset["test"]]).map(lambda x: tokenizer(x[summary_column], truncation=True), batched=True, remove_columns=[text_column, summary_column])
target_lengths = [len(x) for x in tokenized_targets["input_ids"]]
max_target_length = int(np.max(target_lengths))
print(f"Max target length: {max_target_length}")
def preprocess_function(sample, padding="max_length"):
# created prompted input
inputs = [prompt_template.format(input=item) for item in sample[text_column]]
# tokenize inputs
model_inputs = tokenizer(inputs, max_length=max_source_length, padding=padding, truncation=True)
# Tokenize targets with the `text_target` keyword argument
labels = tokenizer(text_target=sample[summary_column], max_length=max_target_length, padding=padding, truncation=True)
# If we are padding here, replace all tokenizer.pad_token_id in the labels by -100 when we want to ignore
# padding in the loss.
if padding == "max_length":
labels["input_ids"] = [
[(l if l != tokenizer.pad_token_id else -100) for l in label] for label in labels["input_ids"]
]
model_inputs["labels"] = labels["input_ids"]
return model_inputs
# process dataset
tokenized_dataset = dataset.map(preprocess_function, batched=True, remove_columns=list(dataset["train"].features))
# save dataset to disk
tokenized_dataset["train"].save_to_disk(os.path.join(save_dataset_path, "train"))
tokenized_dataset["test"].save_to_disk(os.path.join(save_dataset_path, "eval"))