diff options
Diffstat (limited to 'data/textual_inversion')
-rw-r--r-- | data/textual_inversion/csv.py | 11 |
1 files changed, 8 insertions, 3 deletions
diff --git a/data/textual_inversion/csv.py b/data/textual_inversion/csv.py index f306c7a..e082511 100644 --- a/data/textual_inversion/csv.py +++ b/data/textual_inversion/csv.py | |||
@@ -18,7 +18,8 @@ class CSVDataModule(pl.LightningDataModule): | |||
18 | repeats=100, | 18 | repeats=100, |
19 | interpolation="bicubic", | 19 | interpolation="bicubic", |
20 | placeholder_token="*", | 20 | placeholder_token="*", |
21 | center_crop=False): | 21 | center_crop=False, |
22 | valid_set_size=None): | ||
22 | super().__init__() | 23 | super().__init__() |
23 | 24 | ||
24 | self.data_file = Path(data_file) | 25 | self.data_file = Path(data_file) |
@@ -33,6 +34,7 @@ class CSVDataModule(pl.LightningDataModule): | |||
33 | self.placeholder_token = placeholder_token | 34 | self.placeholder_token = placeholder_token |
34 | self.center_crop = center_crop | 35 | self.center_crop = center_crop |
35 | self.interpolation = interpolation | 36 | self.interpolation = interpolation |
37 | self.valid_set_size = valid_set_size | ||
36 | 38 | ||
37 | self.batch_size = batch_size | 39 | self.batch_size = batch_size |
38 | 40 | ||
@@ -44,8 +46,11 @@ class CSVDataModule(pl.LightningDataModule): | |||
44 | self.data_full = [(img, cap) for img, cap, skip in zip(image_paths, captions, skips) if skip != "x"] | 46 | self.data_full = [(img, cap) for img, cap, skip in zip(image_paths, captions, skips) if skip != "x"] |
45 | 47 | ||
46 | def setup(self, stage=None): | 48 | def setup(self, stage=None): |
47 | train_set_size = int(len(self.data_full) * 0.8) | 49 | valid_set_size = int(len(self.data_full) * 0.2) |
48 | valid_set_size = len(self.data_full) - train_set_size | 50 | if self.valid_set_size: |
51 | valid_set_size = math.min(valid_set_size, self.valid_set_size) | ||
52 | train_set_size = len(self.data_full) - valid_set_size | ||
53 | |||
49 | self.data_train, self.data_val = random_split(self.data_full, [train_set_size, valid_set_size]) | 54 | self.data_train, self.data_val = random_split(self.data_full, [train_set_size, valid_set_size]) |
50 | 55 | ||
51 | train_dataset = CSVDataset(self.data_train, self.tokenizer, size=self.size, repeats=self.repeats, interpolation=self.interpolation, | 56 | train_dataset = CSVDataset(self.data_train, self.tokenizer, size=self.size, repeats=self.repeats, interpolation=self.interpolation, |