05f03b541a
* hf offline decorator for tests to workaround rate limits * fail quicker so we can see logs * try new cache name * limit files downloaded * phi mini predownload * offline decorator for phi tokenizer * handle meta llama 8b offline too * make sure to return fixtures if they are wrapped too * more fixes * more things offline * more offline things * fix the env var * fix the model name * handle gemma also * force reload of modules to recheck offline status * prefetch mistral too * use reset_sessions so hub picks up offline mode * more fixes * rename so it doesn't seem like a context manager * fix backoff * switch out tinyshakespeare dataset since it runs a py script to fetch data and doesn't work offline * include additional dataset * more fixes * more fixes * replace tiny shakespeaere dataset * skip some tests for now * use more robust check using snapshot download to determine if a dataset name is on the hub * typo for skip reason * use local_files_only * more fixtures * remove local only * use tiny shakespeare as pretrain dataset and streaming can't be offline even if precached * make sure fixtures aren't offline improve the offline reset try bumping version of datasets reorder reloading and setting prime a new cache run the tests now with fresh cache try with a static cache * now run all the ci again with hopefully a correct cache * skip wonky tests for now * skip wonky tests for now * handle offline mode for model card creation
102 lines
3.2 KiB
Python
102 lines
3.2 KiB
Python
"""Module for testing streaming dataset sequence packing"""
|
|
|
|
import pytest
|
|
from datasets import concatenate_datasets, load_dataset
|
|
from torch.utils.data import DataLoader, RandomSampler
|
|
from transformers import AutoTokenizer
|
|
from utils import enable_hf_offline
|
|
|
|
from axolotl.datasets import TokenizedPromptDataset
|
|
from axolotl.prompt_strategies.completion import load
|
|
from axolotl.utils.collators import V2BatchSamplerDataCollatorForSeq2Seq
|
|
from axolotl.utils.data.utils import drop_long_seq_in_dataset
|
|
from axolotl.utils.dict import DictDefault
|
|
from axolotl.utils.samplers import MultipackBatchSampler, get_dataset_lengths
|
|
|
|
|
|
@pytest.fixture(name="tokenizer")
|
|
def fixture_tokenizer():
|
|
tokenizer = AutoTokenizer.from_pretrained("huggyllama/llama-7b")
|
|
tokenizer.pad_token = "</s>"
|
|
return tokenizer
|
|
|
|
|
|
class TestBatchedSamplerPacking:
|
|
"""
|
|
Test class for packing streaming dataset sequences
|
|
"""
|
|
|
|
@pytest.mark.skip(reason="TODO: fix hf offline mode for CI rate limits")
|
|
@pytest.mark.parametrize(
|
|
"batch_size, num_workers",
|
|
[
|
|
(1, 0),
|
|
(2, 0),
|
|
(1, 2),
|
|
(2, 2),
|
|
],
|
|
)
|
|
@pytest.mark.parametrize("max_seq_length", [4096, 512])
|
|
@enable_hf_offline
|
|
def test_packing(self, batch_size, num_workers, tokenizer, max_seq_length):
|
|
import axolotl.monkeypatch.data.batch_dataset_fetcher # pylint: disable=unused-import # noqa: F401
|
|
|
|
dataset = load_dataset(
|
|
"winglian/tiny-shakespeare",
|
|
split="train",
|
|
)
|
|
|
|
cfg = DictDefault(
|
|
{
|
|
"train_on_inputs": True,
|
|
"sequence_len": max_seq_length,
|
|
}
|
|
)
|
|
ds_cfg = DictDefault(
|
|
{
|
|
"field": "Text",
|
|
}
|
|
)
|
|
completion_strategy = load(tokenizer, cfg, ds_cfg)
|
|
dataset_wrapper = TokenizedPromptDataset(
|
|
completion_strategy,
|
|
dataset,
|
|
)
|
|
train_dataset = concatenate_datasets([dataset_wrapper])
|
|
|
|
train_dataset = drop_long_seq_in_dataset(train_dataset, cfg)
|
|
|
|
lengths = get_dataset_lengths(train_dataset)
|
|
batch_sampler = MultipackBatchSampler(
|
|
sampler=RandomSampler(train_dataset),
|
|
lengths=lengths,
|
|
batch_size=batch_size,
|
|
batch_max_len=max_seq_length,
|
|
group_size=100000,
|
|
bin_size=200,
|
|
)
|
|
|
|
loader = DataLoader(
|
|
train_dataset,
|
|
batch_sampler=batch_sampler,
|
|
collate_fn=V2BatchSamplerDataCollatorForSeq2Seq( # pylint: disable=unexpected-keyword-arg
|
|
tokenizer=tokenizer,
|
|
padding=True,
|
|
pad_to_multiple_of=max_seq_length,
|
|
return_tensors="pt",
|
|
),
|
|
num_workers=num_workers,
|
|
)
|
|
|
|
batch_idxs = []
|
|
for batch in batch_sampler:
|
|
for pack in batch:
|
|
batch_idxs.extend(pack)
|
|
|
|
for batch in loader:
|
|
assert batch["input_ids"].numel() <= batch_size * max_seq_length
|
|
assert batch["input_ids"].shape[1] == max_seq_length
|
|
|
|
original_idxs = set(range(len(train_dataset)))
|
|
assert original_idxs == set(batch_idxs)
|