Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Update datasets to 2.14 #59

Merged
merged 4 commits into from
Jul 28, 2023
Merged

Update datasets to 2.14 #59

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
a0fffa1
fix smoke tests deprecations
LoicGrobol Jul 28, 2023
43a8fc3
bump pyproject
LoicGrobol Jul 28, 2023
d347c5f
add config for ruff to play nice with pytest
LoicGrobol Jul 28, 2023
95d3111
fix remote datasets loading in datasets 2.14
LoicGrobol Jul 28, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 7 additions & 4 deletions pyproject.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ keywords = ["nlp", "transformers", "language-model"]
requires-python = ">=3.8"
dependencies = [
"click >= 8.0.4, < 9.0.0",
"datasets >= 2.2, < 2.14",
"datasets >= 2.2, < 2.15",
"filelock",
"jsonlines",
"loguru",
Expand Down Expand Up @@ -68,6 +68,12 @@ zeldarose-transformer = "zeldarose.train_transformer:main"
line-length = 100
select = ["A", "B", "C90", "E", "F", "N", "NPY", "PT", "PTH", "PYI", "S", "W"]

[tool.ruff.per-file-ignores]
"tests/**/*.py" = [
"S101", # asserts allowed in tests
"ARG", # Unused function args -> fixtures nevertheless are functionally relevant
]


[tool.mypy]
warn_unreachable = true
Expand All @@ -78,6 +84,3 @@ line-length = 100

[tool.isort]
profile = "black"

[tool.bandit]
assert_used.skips = ["tests/test_*.py"]
180 changes: 95 additions & 85 deletions tests/test_smoke.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,14 +25,16 @@ def test_train_tokenizer(
tmp_path: pathlib.Path,
):
ret = script_runner.run(
"zeldarose-tokenizer",
"--vocab-size",
"4096",
"--out-path",
str(tmp_path / "tokenizer"),
"--model-name",
"my-muppet",
str(raw_text_path),
[
"zeldarose-tokenizer",
"--vocab-size",
"4096",
"--out-path",
str(tmp_path / "tokenizer"),
"--model-name",
"my-muppet",
str(raw_text_path),
]
)
assert ret.success

Expand All @@ -57,25 +59,27 @@ def test_train_mbart(
extra_args.extend(["--num-devices", str(devices)])

ret = script_runner.run(
"zeldarose-transformer",
"--accelerator",
accelerator,
"--config",
str(mbart_task_config),
"--tokenizer",
str(mbart_model_config),
"--model-config",
str(mbart_model_config),
"--device-batch-size",
"2",
"--out-dir",
str(tmp_path / "train-out"),
"--cache-dir",
str(tmp_path / "cache"),
"--val-text",
str(translation_dataset_path),
str(translation_dataset_path),
*extra_args,
[
"zeldarose-transformer",
"--accelerator",
accelerator,
"--config",
str(mbart_task_config),
"--tokenizer",
str(mbart_model_config),
"--model-config",
str(mbart_model_config),
"--device-batch-size",
"2",
"--out-dir",
str(tmp_path / "train-out"),
"--cache-dir",
str(tmp_path / "cache"),
"--val-text",
str(translation_dataset_path),
str(translation_dataset_path),
*extra_args,
],
env={"TORCH_DISTRIBUTED_DEBUG": "DETAIL", **os.environ},
)
assert ret.success
Expand All @@ -102,25 +106,27 @@ def test_train_mlm(
extra_args.extend(["--num-devices", str(devices)])

ret = script_runner.run(
"zeldarose-transformer",
"--accelerator",
accelerator,
"--config",
str(mlm_task_config),
"--tokenizer",
str(tokenizer_name_or_path),
"--model-config",
str(mlm_model_config),
"--device-batch-size",
"8",
"--out-dir",
str(tmp_path / "train-out"),
"--cache-dir",
str(tmp_path / "tokenizer-cache"),
"--val-text",
str(raw_text_path),
str(raw_text_path),
*extra_args,
[
"zeldarose-transformer",
"--accelerator",
accelerator,
"--config",
str(mlm_task_config),
"--tokenizer",
str(tokenizer_name_or_path),
"--model-config",
str(mlm_model_config),
"--device-batch-size",
"8",
"--out-dir",
str(tmp_path / "train-out"),
"--cache-dir",
str(tmp_path / "tokenizer-cache"),
"--val-text",
str(raw_text_path),
str(raw_text_path),
*extra_args,
],
env={"TORCH_DISTRIBUTED_DEBUG": "DETAIL", **os.environ},
)
assert ret.success
Expand Down Expand Up @@ -151,25 +157,27 @@ def test_train_rtd(
if num_devices is not None:
extra_args.extend(["--num-devices", str(num_devices)])
ret = script_runner.run(
"zeldarose-transformer",
"--accelerator",
accelerator,
"--config",
str(rtd_task_config),
"--tokenizer",
str(tokenizer_name_or_path),
"--model-config",
str(rtd_model_config),
"--device-batch-size",
"8",
"--out-dir",
str(tmp_path / "train-out"),
"--cache-dir",
str(tmp_path / "tokenizer-cache"),
"--val-text",
str(raw_text_path),
str(raw_text_path),
*extra_args,
[
"zeldarose-transformer",
"--accelerator",
accelerator,
"--config",
str(rtd_task_config),
"--tokenizer",
str(tokenizer_name_or_path),
"--model-config",
str(rtd_model_config),
"--device-batch-size",
"8",
"--out-dir",
str(tmp_path / "train-out"),
"--cache-dir",
str(tmp_path / "tokenizer-cache"),
"--val-text",
str(raw_text_path),
str(raw_text_path),
*extra_args,
],
env={"TORCH_DISTRIBUTED_DEBUG": "DETAIL", **os.environ},
)
assert ret.success
Expand All @@ -182,25 +190,27 @@ def test_train_mlm_with_remote_dataset(
tmp_path: pathlib.Path,
):
ret = script_runner.run(
"zeldarose-transformer",
"--strategy",
"ddp_spawn",
"--num-devices",
"2",
"--config",
str(mlm_task_config),
"--tokenizer",
"lgrobol/roberta-minuscule",
"--model-config",
"lgrobol/roberta-minuscule",
"--device-batch-size",
"8",
"--out-dir",
str(tmp_path / "train-out"),
"--cache-dir",
str(tmp_path / "tokenizer-cache"),
"--val-text",
remote_raw_text,
remote_raw_text,
[
"zeldarose-transformer",
"--strategy",
"ddp_spawn",
"--num-devices",
"2",
"--config",
str(mlm_task_config),
"--tokenizer",
"lgrobol/roberta-minuscule",
"--model-config",
"lgrobol/roberta-minuscule",
"--device-batch-size",
"8",
"--out-dir",
str(tmp_path / "train-out"),
"--cache-dir",
str(tmp_path / "tokenizer-cache"),
"--val-text",
remote_raw_text,
remote_raw_text,
]
)
assert ret.success
5 changes: 4 additions & 1 deletion zeldarose/datasets/transform.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,10 @@ def encode_dataset(
logger.info(f"Loading data from {text_path}")
try:
full_dataset = datasets.load_dataset("text", data_files=str(text_path), split="train")
except FileNotFoundError as e:
# So far the cleaner way to detect that a dataset is remote???
# in datasets < 2.14 this was FileNotFoundError, in 2.14 it's the other one
# in the future? Who's to say,,,
except (FileNotFoundError, datasets.builder.DatasetGenerationError) as e:
if isinstance(text_path, str):
dataset_name, dataset_config, dataset_split = text_path.split(":")
full_dataset = datasets.load_dataset(
Expand Down
Loading