diff options
author | Volpeon <git@volpeon.ink> | 2023-01-02 20:13:59 +0100 |
---|---|---|
committer | Volpeon <git@volpeon.ink> | 2023-01-02 20:13:59 +0100 |
commit | 46d631759f59bc6b65458202641e5f5a9bc30b7b (patch) | |
tree | ea8c94ff336fe27b6cc8f39cea6c1699f44c61d5 /train_dreambooth.py | |
parent | Update (diff) | |
download | textual-inversion-diff-46d631759f59bc6b65458202641e5f5a9bc30b7b.tar.gz textual-inversion-diff-46d631759f59bc6b65458202641e5f5a9bc30b7b.tar.bz2 textual-inversion-diff-46d631759f59bc6b65458202641e5f5a9bc30b7b.zip |
Fixed LR finder
Diffstat (limited to 'train_dreambooth.py')
-rw-r--r-- | train_dreambooth.py | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/train_dreambooth.py b/train_dreambooth.py index 1e49474..218018b 100644 --- a/train_dreambooth.py +++ b/train_dreambooth.py | |||
@@ -634,7 +634,7 @@ def main(): | |||
634 | ) | 634 | ) |
635 | 635 | ||
636 | if args.find_lr: | 636 | if args.find_lr: |
637 | args.learning_rate = 1e2 | 637 | args.learning_rate = 1e-4 |
638 | 638 | ||
639 | # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs | 639 | # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs |
640 | if args.use_8bit_adam: | 640 | if args.use_8bit_adam: |
@@ -901,7 +901,7 @@ def main(): | |||
901 | on_train=lambda: tokenizer.set_use_vector_shuffle(args.vector_shuffle), | 901 | on_train=lambda: tokenizer.set_use_vector_shuffle(args.vector_shuffle), |
902 | on_eval=lambda: tokenizer.set_use_vector_shuffle(False) | 902 | on_eval=lambda: tokenizer.set_use_vector_shuffle(False) |
903 | ) | 903 | ) |
904 | lr_finder.run(min_lr=1e-4) | 904 | lr_finder.run(end_lr=1e2) |
905 | 905 | ||
906 | plt.savefig(basepath.joinpath("lr.png")) | 906 | plt.savefig(basepath.joinpath("lr.png")) |
907 | plt.close() | 907 | plt.close() |