File size: 3,503 Bytes
98ced8b
 
968f4bc
98ced8b
 
 
 
 
 
968f4bc
98ced8b
 
968f4bc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
98ced8b
 
 
 
 
053730f
98ced8b
 
3a7ead3
8382f82
98ced8b
 
 
8382f82
968f4bc
98ced8b
053730f
159baa9
 
 
 
98ced8b
 
 
3a7ead3
 
 
 
98ced8b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
8382f82
 
98ced8b
3a7ead3
98ced8b
0cde3e9
 
98ced8b
 
 
 
 
 
968f4bc
98ced8b
159baa9
 
 
 
 
968f4bc
 
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
import evaluate
import numpy as np
import streamlit as st
from datasets import load_dataset
from transformers import (
    AutoModelForSequenceClassification,
    AutoTokenizer,
    DataCollatorWithPadding,
    Trainer,
    TrainerCallback,
    TrainingArguments,
)
from transformers.trainer_callback import TrainerControl, TrainerState

import wandb


class StreamlitProgressbarCallback(TrainerCallback):

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.progress_bar = st.progress(0, text="Training")

    def on_step_begin(
        self,
        args: TrainingArguments,
        state: TrainerState,
        control: TrainerControl,
        **kwargs,
    ):
        super().on_step_begin(args, state, control, **kwargs)
        self.progress_bar.progress(
            (state.global_step * 100 // state.max_steps) + 1,
            text=f"Training {state.global_step} / {state.max_steps}",
        )


def train_binary_classifier(
    project_name: str,
    entity_name: str,
    run_name: str,
    dataset_repo: str = "geekyrakshit/prompt-injection-dataset",
    model_name: str = "distilbert/distilbert-base-uncased",
    prompt_column_name: str = "prompt",
    learning_rate: float = 1e-5,
    batch_size: int = 16,
    num_epochs: int = 2,
    weight_decay: float = 0.01,
    save_steps: int = 1000,
    streamlit_mode: bool = False,
):
    wandb.init(project=project_name, entity=entity_name, name=run_name)
    if streamlit_mode:
        st.markdown(
            f"Explore your training logs on [Weights & Biases]({wandb.run.url})"
        )
    dataset = load_dataset(dataset_repo)
    tokenizer = AutoTokenizer.from_pretrained(model_name)

    tokenized_datasets = dataset.map(
        lambda examples: tokenizer(examples[prompt_column_name], truncation=True),
        batched=True,
    )
    data_collator = DataCollatorWithPadding(tokenizer=tokenizer)
    accuracy = evaluate.load("accuracy")

    def compute_metrics(eval_pred):
        predictions, labels = eval_pred
        predictions = np.argmax(predictions, axis=1)
        return accuracy.compute(predictions=predictions, references=labels)

    id2label = {0: "SAFE", 1: "INJECTION"}
    label2id = {"SAFE": 0, "INJECTION": 1}

    model = AutoModelForSequenceClassification.from_pretrained(
        model_name,
        num_labels=2,
        id2label=id2label,
        label2id=label2id,
    )

    trainer = Trainer(
        model=model,
        args=TrainingArguments(
            output_dir="binary-classifier",
            learning_rate=learning_rate,
            per_device_train_batch_size=batch_size,
            per_device_eval_batch_size=batch_size,
            num_train_epochs=num_epochs,
            weight_decay=weight_decay,
            eval_strategy="epoch",
            save_strategy="steps",
            save_steps=save_steps,
            load_best_model_at_end=True,
            push_to_hub=False,
            report_to="wandb",
            logging_strategy="steps",
            logging_steps=1,
        ),
        train_dataset=tokenized_datasets["train"],
        eval_dataset=tokenized_datasets["test"],
        processing_class=tokenizer,
        data_collator=data_collator,
        compute_metrics=compute_metrics,
        callbacks=[StreamlitProgressbarCallback()] if streamlit_mode else [],
    )
    try:
        training_output = trainer.train()
    except Exception as e:
        wandb.finish()
        raise e
    wandb.finish()
    return training_output