diff options
| author | Marcin Chrzanowski <m@m-chrzan.xyz> | 2021-05-25 17:36:02 +0200 |
|---|---|---|
| committer | Marcin Chrzanowski <m@m-chrzan.xyz> | 2021-05-25 17:36:02 +0200 |
| commit | a2dd5a009a73d1ea5f03894320af818c05c5778d (patch) | |
| tree | a53abc8405c04a0cd4f0f31401f0cc7cff7dbdd7 /data | |
| parent | 0b67d2c000f31c9048c2ea346448184f1ce97e0d (diff) | |
Allow for configurable test set parameters
Diffstat (limited to 'data')
| -rw-r--r-- | data/testset.py | 5 |
1 files changed, 2 insertions, 3 deletions
diff --git a/data/testset.py b/data/testset.py index 07fc811..f8b938e 100644 --- a/data/testset.py +++ b/data/testset.py @@ -4,9 +4,8 @@ from data.generate import get_single_example TEST_SIZE = 128 -test_examples = [get_single_example() for i in range(TEST_SIZE)] - -def get_testset(device='cpu'): +def get_testset(n_tokens=16, seqlen=64, max_count=9, device='cpu'): + test_examples = [get_single_example(n_tokens, seqlen, max_count) for i in range(TEST_SIZE)] # Transpositions are used, because the convention in PyTorch is to represent # sequence tensors as <seq_len, batch_size> instead of <batch_size, seq_len>. test_X = torch.tensor( |