File size: 9,833 Bytes
e634118 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 |
"""
Test dataset loading under various conditions.
"""
import shutil
import tempfile
import unittest
from pathlib import Path
from datasets import Dataset
from huggingface_hub import snapshot_download
from transformers import AutoTokenizer
from axolotl.utils.data import load_tokenized_prepared_datasets
from axolotl.utils.dict import DictDefault
class TestDatasetPreparation(unittest.TestCase):
"""Test a configured dataloader."""
def setUp(self) -> None:
self.tokenizer = AutoTokenizer.from_pretrained("huggyllama/llama-7b")
self.tokenizer.add_special_tokens(
{
"bos_token": "<s>",
"eos_token": "</s>",
"unk_token": "<unk>",
}
)
# Alpaca dataset.
self.dataset = Dataset.from_list(
[
{
"instruction": "Evaluate this sentence for spelling and grammar mistakes",
"input": "He finnished his meal and left the resturant",
"output": "He finished his meal and left the restaurant.",
}
]
)
def test_load_hub(self):
"""Core use case. Verify that processing data from the hub works"""
with tempfile.TemporaryDirectory() as tmp_dir:
prepared_path = Path(tmp_dir) / "prepared"
cfg = DictDefault(
{
"tokenizer_config": "huggyllama/llama-7b",
"sequence_len": 1024,
"datasets": [
{
"path": "mhenrichsen/alpaca_2k_test",
"type": "alpaca",
},
],
}
)
dataset, _ = load_tokenized_prepared_datasets(
self.tokenizer, cfg, prepared_path
)
assert len(dataset) == 2000
assert "input_ids" in dataset.features
assert "attention_mask" in dataset.features
assert "labels" in dataset.features
def test_load_local_hub(self):
"""Niche use case. Verify that a local copy of a hub dataset can be loaded"""
with tempfile.TemporaryDirectory() as tmp_dir:
tmp_ds_path = Path("mhenrichsen/alpaca_2k_test")
tmp_ds_path.mkdir(parents=True, exist_ok=True)
snapshot_download(
repo_id="mhenrichsen/alpaca_2k_test",
repo_type="dataset",
local_dir=tmp_ds_path,
)
prepared_path = Path(tmp_dir) / "prepared"
# Right now a local copy that doesn't fully conform to a dataset
# must list data_files and ds_type otherwise the loader won't know
# how to load it.
cfg = DictDefault(
{
"tokenizer_config": "huggyllama/llama-7b",
"sequence_len": 1024,
"datasets": [
{
"path": "mhenrichsen/alpaca_2k_test",
"ds_type": "parquet",
"type": "alpaca",
"data_files": [
"mhenrichsen/alpaca_2k_test/alpaca_2000.parquet",
],
},
],
}
)
dataset, _ = load_tokenized_prepared_datasets(
self.tokenizer, cfg, prepared_path
)
assert len(dataset) == 2000
assert "input_ids" in dataset.features
assert "attention_mask" in dataset.features
assert "labels" in dataset.features
shutil.rmtree(tmp_ds_path)
def test_load_from_save_to_disk(self):
"""Usual use case. Verify datasets saved via `save_to_disk` can be loaded."""
with tempfile.TemporaryDirectory() as tmp_dir:
tmp_ds_name = Path(tmp_dir) / "tmp_dataset"
self.dataset.save_to_disk(tmp_ds_name)
prepared_path = Path(tmp_dir) / "prepared"
cfg = DictDefault(
{
"tokenizer_config": "huggyllama/llama-7b",
"sequence_len": 256,
"datasets": [
{
"path": str(tmp_ds_name),
"type": "alpaca",
},
],
}
)
dataset, _ = load_tokenized_prepared_datasets(
self.tokenizer, cfg, prepared_path
)
assert len(dataset) == 1
assert "input_ids" in dataset.features
assert "attention_mask" in dataset.features
assert "labels" in dataset.features
def test_load_from_dir_of_parquet(self):
"""Usual use case. Verify a directory of parquet files can be loaded."""
with tempfile.TemporaryDirectory() as tmp_dir:
tmp_ds_dir = Path(tmp_dir) / "tmp_dataset"
tmp_ds_dir.mkdir()
tmp_ds_path = tmp_ds_dir / "shard1.parquet"
self.dataset.to_parquet(tmp_ds_path)
prepared_path: Path = Path(tmp_dir) / "prepared"
cfg = DictDefault(
{
"tokenizer_config": "huggyllama/llama-7b",
"sequence_len": 256,
"datasets": [
{
"path": str(tmp_ds_dir),
"ds_type": "parquet",
"name": "test_data",
"data_files": [
str(tmp_ds_path),
],
"type": "alpaca",
},
],
}
)
dataset, _ = load_tokenized_prepared_datasets(
self.tokenizer, cfg, prepared_path
)
assert len(dataset) == 1
assert "input_ids" in dataset.features
assert "attention_mask" in dataset.features
assert "labels" in dataset.features
def test_load_from_dir_of_json(self):
"""Standard use case. Verify a directory of json files can be loaded."""
with tempfile.TemporaryDirectory() as tmp_dir:
tmp_ds_dir = Path(tmp_dir) / "tmp_dataset"
tmp_ds_dir.mkdir()
tmp_ds_path = tmp_ds_dir / "shard1.json"
self.dataset.to_json(tmp_ds_path)
prepared_path: Path = Path(tmp_dir) / "prepared"
cfg = DictDefault(
{
"tokenizer_config": "huggyllama/llama-7b",
"sequence_len": 256,
"datasets": [
{
"path": str(tmp_ds_dir),
"ds_type": "json",
"name": "test_data",
"data_files": [
str(tmp_ds_path),
],
"type": "alpaca",
},
],
}
)
dataset, _ = load_tokenized_prepared_datasets(
self.tokenizer, cfg, prepared_path
)
assert len(dataset) == 1
assert "input_ids" in dataset.features
assert "attention_mask" in dataset.features
assert "labels" in dataset.features
def test_load_from_single_parquet(self):
"""Standard use case. Verify a single parquet file can be loaded."""
with tempfile.TemporaryDirectory() as tmp_dir:
tmp_ds_path = Path(tmp_dir) / "tmp_dataset.parquet"
self.dataset.to_parquet(tmp_ds_path)
prepared_path: Path = Path(tmp_dir) / "prepared"
cfg = DictDefault(
{
"tokenizer_config": "huggyllama/llama-7b",
"sequence_len": 256,
"datasets": [
{
"path": str(tmp_ds_path),
"name": "test_data",
"type": "alpaca",
},
],
}
)
dataset, _ = load_tokenized_prepared_datasets(
self.tokenizer, cfg, prepared_path
)
assert len(dataset) == 1
assert "input_ids" in dataset.features
assert "attention_mask" in dataset.features
assert "labels" in dataset.features
def test_load_from_single_json(self):
"""Standard use case. Verify a single json file can be loaded."""
with tempfile.TemporaryDirectory() as tmp_dir:
tmp_ds_path = Path(tmp_dir) / "tmp_dataset.json"
self.dataset.to_json(tmp_ds_path)
prepared_path: Path = Path(tmp_dir) / "prepared"
cfg = DictDefault(
{
"tokenizer_config": "huggyllama/llama-7b",
"sequence_len": 256,
"datasets": [
{
"path": str(tmp_ds_path),
"name": "test_data",
"type": "alpaca",
},
],
}
)
dataset, _ = load_tokenized_prepared_datasets(
self.tokenizer, cfg, prepared_path
)
assert len(dataset) == 1
assert "input_ids" in dataset.features
assert "attention_mask" in dataset.features
assert "labels" in dataset.features
if __name__ == "__main__":
unittest.main()
|