-
Notifications
You must be signed in to change notification settings - Fork 159
/
multitask_train.py
132 lines (111 loc) · 4.44 KB
/
multitask_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
# -*- coding: utf-8 -*-
# file: train.py
# time: 11:30 2023/3/13
# author: YANG, HENG <hy345@exeter.ac.uk> (杨恒)
# github: https://github.com/yangheng95
# huggingface: https://huggingface.co/yangheng
# google scholar: https://scholar.google.com/citations?user=NPq5a_0AAAAJ&hl=en
# Copyright (C) 2019-2023. All Rights Reserved.
import os
import warnings
warnings.filterwarnings("ignore")
import pandas as pd
from .model import T5Generator, T5Classifier
from .data_utils import InstructDatasetLoader, read_json
task_name = "multitask"
experiment_name = "instruction"
# model_checkpoint = 'allenai/tk-instruct-base-def-pos'
model_checkpoint = "kevinscaria/ate_tk-instruct-base-def-pos-neg-neut-combined"
# model_checkpoint = 'allenai/tk-instruct-large-def-pos'
# model_checkpoint = 'allenai/tk-instruct-3b-def-pos'
# model_checkpoint = 'google/mt5-base'
print("Experiment Name: ", experiment_name)
model_out_path = "checkpoints"
model_out_path = os.path.join(
model_out_path, task_name, f"{model_checkpoint.replace('/', '')}-{experiment_name}"
)
print("Model output path: ", model_out_path)
# Load the data
# id_train_file_path = './integrated_datasets'
# id_test_file_path = './integrated_datasets'
id_train_file_path = "./integrated_datasets/acos_datasets/"
id_test_file_path = "./integrated_datasets/acos_datasets"
# id_train_file_path = './integrated_datasets/acos_datasets/501.Laptop14'
# id_test_file_path = './integrated_datasets/acos_datasets/501.Laptop14'
# id_train_file_path = './integrated_datasets/acos_datasets/504.Restaurant16'
# id_test_file_path = './integrated_datasets/acos_datasets/504.Restaurant16'
id_tr_df = read_json(id_train_file_path, "train")
id_te_df = read_json(id_test_file_path, "test")
id_tr_df = pd.DataFrame(id_tr_df)
id_te_df = pd.DataFrame(id_te_df)
loader = InstructDatasetLoader(id_tr_df, id_te_df)
if loader.train_df_id is not None:
loader.train_df_id = loader.prepare_instruction_dataloader(loader.train_df_id)
if loader.test_df_id is not None:
loader.test_df_id = loader.prepare_instruction_dataloader(loader.test_df_id)
if loader.train_df_ood is not None:
loader.train_df_ood = loader.prepare_instruction_dataloader(loader.train_df_ood)
if loader.test_df_ood is not None:
loader.test_df_ood = loader.prepare_instruction_dataloader(loader.test_df_ood)
# Create T5 utils object
t5_exp = T5Generator(model_checkpoint)
# Tokenize Dataset
id_ds, id_tokenized_ds, ood_ds, ood_tokenzed_ds = loader.create_datasets(
t5_exp.tokenize_function_inputs
)
# Training arguments
training_args = {
"output_dir": model_out_path,
"evaluation_strategy": "epoch",
"save_strategy": "epoch",
"learning_rate": 5e-5,
"per_device_train_batch_size": 16,
"per_device_eval_batch_size": 16,
"num_train_epochs": 6,
"weight_decay": 0.01,
"warmup_ratio": 0.1,
"load_best_model_at_end": True,
"push_to_hub": False,
"eval_accumulation_steps": 1,
"predict_with_generate": True,
"logging_steps": 1000000000,
"use_mps_device": False,
# 'fp16': True,
"fp16": False,
}
# Train model
model_trainer = t5_exp.train(id_tokenized_ds, **training_args)
# Model inference - Trainer object - (Pass model trainer as predictor)
# model_checkpoint = findfile.find_cwd_dir('tk-instruct-base-def-pos')
# t5_exp = T5Generator(model_checkpoint)
# Get prediction labels - Training set
id_tr_pred_labels = t5_exp.get_labels(
predictor=model_trainer,
tokenized_dataset=id_tokenized_ds,
sample_set="train",
batch_size=16,
)
id_tr_labels = [i.strip() for i in id_ds["train"]["labels"]]
# Get prediction labels - Testing set
id_te_pred_labels = t5_exp.get_labels(
predictor=model_trainer,
tokenized_dataset=id_tokenized_ds,
sample_set="test",
batch_size=16,
)
id_te_labels = [i.strip() for i in id_ds["test"]["labels"]]
# # Compute Metrics
# metrics = t5_exp.get_metrics(id_tr_labels, id_tr_pred_labels)
# print('----------------------- Training Set Metrics -----------------------')
# print(metrics)
#
# metrics = t5_exp.get_metrics(id_te_labels, id_te_pred_labels)
# print('----------------------- Testing Set Metrics -----------------------')
# print(metrics)
# Compute Metrics
metrics = t5_exp.get_classic_metrics(id_tr_labels, id_tr_pred_labels)
print("----------------------- Classic Training Set Metrics -----------------------")
print(metrics)
metrics = t5_exp.get_classic_metrics(id_te_labels, id_te_pred_labels)
print("----------------------- Classic Testing Set Metrics -----------------------")
print(metrics)