QuantEase/datautils.py at main · linkedin/QuantEase · GitHub

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
# Codes are taken from https://github.com/IST-DASLab/gptq with modifications

import json
import os

import numpy as np
import torch

ROOT_DIR = "."

os.environ["HF_DATASETS_OFFLINE"] = "true"


def set_seed(seed):
    np.random.seed(seed)
    torch.random.manual_seed(seed)


def get_wikitext(nsamples, seed, seqlen, model, cached=True, read_json=True):
    if not cached:
        from datasets import load_dataset, load_from_disk
        traindata = load_dataset('wikitext', 'wikitext-2-raw-v1', split='train',
                                 cache_dir=os.path.join(ROOT_DIR, 'data'))
        testdata = load_dataset('wikitext', 'wikitext-2-raw-v1', split='test', cache_dir=os.path.join(ROOT_DIR, 'data'))

        traindata.save_to_disk(os.path.join(ROOT_DIR, 'data', 'wikitext', 'wiki-train'))
        testdata.save_to_disk(os.path.join(ROOT_DIR, 'data', 'wikitext', 'wiki-test'))
    elif read_json is False:
        from datasets import load_dataset, load_from_disk
        traindata = load_from_disk(os.path.join(ROOT_DIR, 'data', 'wikitext', 'wiki-train'))
        testdata = load_from_disk(os.path.join(ROOT_DIR, 'data', 'wikitext', 'wiki-test'))
        traindata = traindata['text']
        testdata = testdata['text']
    else:
        f = open(os.path.join(ROOT_DIR, 'data', 'wikitrain.json'))
        traindata = json.load(f)
        f = open(os.path.join(ROOT_DIR, 'data', 'wikitest.json'))
        testdata = json.load(f)

    from transformers import AutoTokenizer
    tokenizer = AutoTokenizer.from_pretrained(model, use_fast=False, local_files_only=cached,
                                              cache_dir=os.path.join(ROOT_DIR, 'data'), trust_remote_code=True,
                                              use_auth_token=False)
    trainenc = tokenizer("\n\n".join(traindata), return_tensors='pt')
    testenc = tokenizer("\n\n".join(testdata), return_tensors='pt')

    import random
    random.seed(seed)
    trainloader = []
    for _ in range(nsamples):
        i = random.randint(0, trainenc.input_ids.shape[1] - seqlen - 1)
        j = i + seqlen
        inp = trainenc.input_ids[:, i:j]
        tar = inp.clone()
        tar[:, :-1] = -100
        trainloader.append((inp, tar))
    return trainloader, testenc


def get_ptb(nsamples, seed, seqlen, model, cached=True, read_json=True):
    if not cached:
        from datasets import load_dataset, load_from_disk
        traindata = load_dataset('ptb_text_only', 'penn_treebank', split='train',
                                 cache_dir=os.path.join(ROOT_DIR, 'data'))
        valdata = load_dataset('ptb_text_only', 'penn_treebank', split='validation',
                               cache_dir=os.path.join(ROOT_DIR, 'data'))
        traindata.save_to_disk(os.path.join(ROOT_DIR, 'data', 'ptb', 'ptb-train'))
        valdata.save_to_disk(os.path.join(ROOT_DIR, 'data', 'ptb', 'ptb-val'))
    elif not read_json:
        from datasets import load_dataset, load_from_disk
        traindata = load_from_disk(os.path.join(ROOT_DIR, 'data', 'ptb', 'ptb-train'))
        valdata = load_from_disk(os.path.join(ROOT_DIR, 'data', 'ptb', 'ptb-val'))
        traindata = traindata['sentence']
        valdata = valdata['sentence']
    else:
        f = open(os.path.join(ROOT_DIR, 'data', 'ptbtrain.json'))
        traindata = json.load(f)
        f = open(os.path.join(ROOT_DIR, 'data', 'ptbval.json'))
        valdata = json.load(f)

    from transformers import AutoTokenizer
    tokenizer = AutoTokenizer.from_pretrained(model, use_fast=False, local_files_only=cached,
                                              cache_dir=os.path.join(ROOT_DIR, 'data'), trust_remote_code=True)
    trainenc = tokenizer("\n\n".join(traindata), return_tensors='pt')
    testenc = tokenizer("\n\n".join(valdata), return_tensors='pt')

    import random
    random.seed(seed)
    trainloader = []
    for _ in range(nsamples):
        i = random.randint(0, trainenc.input_ids.shape[1] - seqlen - 1)
        j = i + seqlen
        inp = trainenc.input_ids[:, i:j]
        tar = inp.clone()
        tar[:, :-1] = -100
        trainloader.append((inp, tar))
    return trainloader, testenc


def get_c4_val(nsamples, seed, seqlen, model, cached=True):
    if not cached:
        from datasets import load_dataset
        valdata = load_dataset(
            'allenai/c4', 'allenai--c4', data_files={'validation': 'en/c4-validation.00000-of-00008.json.gz'},
            split='validation',
            cache_dir=os.path.join(ROOT_DIR, 'data')
        )
        valdata.save_to_disk(os.path.join(ROOT_DIR, 'data', 'c4', 'c4-val'))
    else:
        from datasets import load_from_disk
        valdata = load_from_disk(os.path.join(ROOT_DIR, 'data', 'c4', 'c4-val'))

    from transformers import AutoTokenizer
    print(model)
    print(os.path.join(ROOT_DIR, 'data'))
    tokenizer = AutoTokenizer.from_pretrained(model, use_fast=False, trust_remote_code=True)
    import random
    random.seed(seed)
    trainloader = []
    for jjj in range(nsamples):
        while True:
            i = random.randint(0, len(valdata) - 1)
            trainenc = tokenizer(valdata[i]['text'], return_tensors='pt')
            if trainenc.input_ids.shape[1] >= seqlen + 1:
                break
        i = random.randint(0, trainenc.input_ids.shape[1] - seqlen - 1)
        j = i + seqlen
        inp = trainenc.input_ids[:, i:j]
        tar = inp.clone()
        tar[:, :-1] = -100
        trainloader.append((inp, tar))

    import random

    # valenc = tokenizer("\n\n".join(valdata['text']), return_tensors='pt')
    random.seed(0)
    valenc = []
    for _ in range(256):
        while True:
            i = random.randint(0, len(valdata) - 1)
            tmp = tokenizer(valdata[i]['text'], return_tensors='pt')
            if tmp.input_ids.shape[1] >= seqlen:
                break
        i = random.randint(0, tmp.input_ids.shape[1] - seqlen - 1)
        j = i + seqlen
        valenc.append(tmp.input_ids[:, i:j])
    valenc = torch.hstack(valenc)

    class TokenizerWrapper:
        def __init__(self, input_ids):
            self.input_ids = input_ids

    valenc = TokenizerWrapper(valenc)

    return trainloader, valenc


def get_loaders(name, nsamples=128, seed=0, seqlen=2048, model=''):
    if 'wikitext2' in name:
        return get_wikitext(nsamples, seed, seqlen, model)
    if 'ptb' in name:
        return get_ptb(nsamples, seed, seqlen, model)
    if 'c4' in name:
        return get_c4_val(nsamples, seed, seqlen, model)