forked from phenixace/TOMG-Bench
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathtest.py
190 lines (157 loc) · 8.26 KB
/
test.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
'''
For fine-tuning inference
'''
import os
import re
import json
import random
import rdkit
import torch
import argparse
import transformers
import pandas as pd
from tqdm import tqdm
from transformers import AutoTokenizer, AutoModelForCausalLM, T5ForConditionalGeneration, GenerationConfig
from utils.dataset import OMGDataset, TMGDataset, OMGInsTDataset
from peft import (
PeftModel
)
def main():
parser = argparse.ArgumentParser()
parser.add_argument("--model_type", type=str, default="t5")
parser.add_argument("--name", type=str, default="molt5-large")
parser.add_argument("--base_model", type=str, default="laituan245/molt5-large-caption2smiles") # meta-llama/Meta-Llama-3-8B-Instruct
parser.add_argument("--adapter_path", type=str, default="laituan245/molt5-large-caption2smiles")
parser.add_argument("--benchmark", type=str, default="open_generation")
parser.add_argument("--task", type=str, default="MolCustom")
parser.add_argument("--subtask", type=str, default="AtomNum")
parser.add_argument("--output_dir", type=str, default="./new_predictions/")
parser.add_argument("--cutoff_len", type=int, default=768) # anyway, reserve 256 tokens for generation (batch infer)
# generation config
parser.add_argument("--temperature", type=float, default=0.75)
parser.add_argument("--top_p", type=float, default=0.85)
parser.add_argument("--top_k", type=int, default=40)
parser.add_argument("--num_beams", type=int, default=1)
parser.add_argument("--num_return_sequences", type=int, default=1)
parser.add_argument("--max_new_tokens", type=int, default=128)
parser.add_argument("--seed", type=int, default=42)
# partition inference
parser.add_argument("--partition", type=int, default=1)
parser.add_argument("--cur", type=int, default=1)
parser.add_argument("--enable_lora", default=False, action="store_true")
parser.add_argument("--int4", default=False, action="store_true")
parser.add_argument("--int8", default=False, action="store_true")
parser.add_argument("--fp16", default=False, action="store_true")
# parser.add_argument("--bf16", default=False, action="store_true") # in case someone wants to use the bf16 option
parser.add_argument("--selfies", default=False, action="store_true")
parser.add_argument("--smiles_check", default=False, action="store_true")
args = parser.parse_args()
# check out put dir
args.output_dir = args.output_dir + "/" + args.name + "/" + args.benchmark + "/" + args.task + "/"
if not os.path.exists(args.output_dir):
os.makedirs(args.output_dir)
if os.path.exists(args.output_dir + "/" + args.subtask + "_" + str(args.cur) + ".csv"):
temp = pd.read_csv(args.output_dir + "/" + args.subtask + "_" + str(args.cur) + ".csv")
start_pos = len(temp)
else:
with open(args.output_dir + args.subtask + "_" + str(args.cur) + ".csv", "w+") as f:
f.write("outputs\n")
start_pos = 0
print("Start from: ", start_pos)
# set random seeds
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
torch.cuda.manual_seed_all(args.seed)
transformers.set_seed(args.seed)
random.seed(args.seed)
# print parameters
print("========Parameters========")
for attr, value in args.__dict__.items():
print("{}={}".format(attr.upper(), value))
# load dataset
if args.benchmark == "open_generation":
if args.model_type == "t5":
if args.selfies:
test_dataset = OMGDataset(args.task, args.subtask, use_selfies=True)
else:
test_dataset = OMGDataset(args.task, args.subtask)
else:
test_dataset = OMGInsTDataset(args.task, args.subtask)
elif args.benchmark == "targeted_generation":
test_dataset = TMGDataset(args.task, args.subtask)
else:
raise ValueError("Invalid benchmark: {}".format(args.benchmark))
# load tokenizer
tokenizer = AutoTokenizer.from_pretrained(args.base_model)
# load model
device_map = "auto"
if args.enable_lora:
model = AutoModelForCausalLM.from_pretrained(args.base_model, load_in_8bit=True if args.int8 else False, load_in_4bit=True if args.int4 else False, torch_dtype=torch.float16 if args.fp16 else torch.float32, device_map=device_map)
model = PeftModel.from_pretrained(model, args.adapter_path, torch_dtype=torch.float16 if args.fp16 else torch.float32, device_map=device_map)
else:
if args.model_type == "decoder-only":
model = AutoModelForCausalLM.from_pretrained(args.adapter_path, load_in_4bit=True if args.int4 else False, load_in_8bit=True if args.int8 else False, torch_dtype=torch.float16 if args.fp16 else torch.float32, device_map=device_map)
elif args.model_type == "t5": # for molt5 and biot5
model = T5ForConditionalGeneration.from_pretrained(args.adapter_path, device_map=device_map)
model.half()
model.eval()
generation_config = GenerationConfig(
do_sample=True,
temperature=args.temperature,
top_p=args.top_p,
top_k=args.top_k,
num_beams=args.num_beams,
pad_token_id=0,
)
# evaluate
with tqdm(total=int(len(test_dataset)*args.cur/args.partition)-int(len(test_dataset)*(args.cur-1)/args.partition)-start_pos) as pbar:
pbar.set_description("Inference")
for idx in range(int(len(test_dataset)*(args.cur-1)/args.partition)+start_pos, int(len(test_dataset)*args.cur/args.partition)):
error_count = 0
while True:
if args.model_type == "t5":
text_input = test_dataset.instructions[idx]
else:
text_input = test_dataset.data[idx]
print(text_input)
model_input = tokenizer(text_input, return_tensors="pt")["input_ids"].cuda()
# labels = tokenizer(test_data[idx][1], return_tensors="pt")
with torch.no_grad():
# print(model_input)
generation = model.generate(
inputs=model_input,
generation_config=generation_config,
return_dict_in_generate=True,
output_scores=True,
max_new_tokens=args.max_new_tokens,
num_return_sequences=args.num_return_sequences
)
if args.subtask in ["bbbp-uni", "bace-uni", "clintox-uni", "hiv-uni", "toxcast-uni", "sider-uni", "tox21-uni"]:
yes_token_id = tokenizer.convert_tokens_to_ids("▁Yes") # mistral
no_token_id = tokenizer.convert_tokens_to_ids("▁No")
yes_token_id = tokenizer.convert_tokens_to_ids("ĠYes") # llama3
no_token_id = tokenizer.convert_tokens_to_ids("ĠNo")
s = tokenizer.decode(generation.sequences[0], skip_special_tokens=True)
scores = generation.scores[0].softmax(dim=-1)
logits = torch.tensor(scores[:,[yes_token_id, no_token_id]], dtype=torch.float32).softmax(dim=-1)[0]
s += "\t" + str(logits[0].item())
else:
s = tokenizer.decode(generation.sequences[0], skip_special_tokens=True)
if args.model_type == "decoder_only":
s = s.split("## Molecule:")[1]
if args.smiles_check:
try:
mol = rdkit.Chem.MolFromSmiles(s.strip().strip("\n").strip())
except:
error_count += 1
print("Error: ", s + " can not be converted to mol")
if error_count > 10:
break
else:
break
print(s)
df = pd.DataFrame([s.strip()], columns=["outputs"])
df.to_csv(args.output_dir + "/" + args.subtask + "_" + str(args.cur) + ".csv", mode='a', header=False, index=True)
pbar.update(1)
if __name__ == "__main__":
main()