Skip to content

Fix output transform, add test to enforce tokenizer consistency #73

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

Merged
lostella merged 2 commits into from
May 17, 2024
+40 −2
Merged

Fix output transform, add test to enforce tokenizer consistency #73

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
65006ba
Fix bin indexing, add tokenizer consistency test
HugoSenetaire May 17, 2024
7864764
fix : black formatting
HugoSenetaire May 17, 2024
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
2 changes: 1 addition & 1 deletion src/chronos/chronos.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -185,7 +185,7 @@ def output_transform(
) -> torch.Tensor:
scale_unsqueezed = scale.unsqueeze(-1).unsqueeze(-1)
indices = torch.clamp(
samples - self.config.n_special_tokens,
samples - self.config.n_special_tokens - 1,
min=0,
max=len(self.centers) - 1,
)
Expand Down
40 changes: 39 additions & 1 deletion test/test_chronos.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,45 @@
import torch
import pytest

from chronos import ChronosConfig, ChronosPipeline
from chronos import ChronosConfig, ChronosPipeline, MeanScaleUniformBins


@pytest.mark.parametrize("n_numerical_tokens", [5, 10, 27])
@pytest.mark.parametrize("n_special_tokens", [2, 5, 13])
def test_tokenizer_consistency(n_numerical_tokens: int, n_special_tokens: int):
n_tokens = n_numerical_tokens + n_special_tokens

config = ChronosConfig(
tokenizer_class="MeanScaleUniformBins",
tokenizer_kwargs=dict(low_limit=-1.0, high_limit=1.0),
n_tokens=n_tokens,
n_special_tokens=n_special_tokens,
pad_token_id=0,
eos_token_id=1,
use_eos_token=True,
model_type="seq2seq",
context_length=512,
prediction_length=64,
num_samples=20,
temperature=1.0,
top_k=50,
top_p=1.0,
)

tokenizer = config.create_tokenizer()
assert isinstance(tokenizer, MeanScaleUniformBins)

context = tokenizer.centers.unsqueeze(0) # add batch dimension
scale = torch.ones((1,)) # fix the scale to one to turn off scaling

token_ids, _, _ = tokenizer.input_transform(context, scale=scale)

samples = tokenizer.output_transform(
token_ids[:, :-1].unsqueeze(1), # remove final EOS, add sample dimension
scale=scale,
)

assert (samples[0, 0, :] == context).all()


@pytest.mark.xfail
Expand Down