Skip to content

Commit 56cb383

Browse files
authored
Actually fix links to pretrained models.
1 parent 7fac857 commit 56cb383

File tree

1 file changed

+11
-1
lines changed

1 file changed

+11
-1
lines changed

README.md

Lines changed: 11 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -93,10 +93,20 @@ python3 compress.py -i path/to/image/dir -ckpt path/to/trained/model
9393

9494
### Pretrained Models
9595

96-
* Pretrained models using the OpenImages dataset can be found below. The examples at the end of this readme were produced using the `HIFIC-med` model. Each model was trained for around `2e5` warmup steps and `2e5` steps with the full generative loss, with a target bitrate of `bpp={'low': 0.14, 'med': 0.3, 'high': 0.45}`. To use a pretrained model, download the model (around 2 GB) and point the `-ckpt` argument in the command above to the corresponding path. If you want to finetune this model, e.g. on some domain-specific dataset, use the following options (you will probably need to adapt the learning rate schedule yourself):
96+
* Pretrained models using the OpenImages dataset can be found below. The examples at the end of this readme were produced using the `HIFIC-med` model. Each model was trained for around `2e5` warmup steps and `2e5` steps with the full generative loss, with a target bitrate of `bpp={'low': 0.14, 'med': 0.3, 'high': 0.45}`. Note the original paper trained for `1e6` steps in each mode, so you can probably get better performance by training from scratch yourself.
97+
98+
* To use a pretrained model, download the model (around 2 GB) and point the `-ckpt` argument in the command above to the corresponding path. If you want to finetune this model, e.g. on some domain-specific dataset, use the following options for each respective model (you will probably need to adapt the learning rate and rate-penalty schedule yourself):
9799

98100
```bash
101+
# Low regime
102+
python3 train.py --model_type compression_gan --regime low --warmstart -ckpt path/to/trained/model -nrb 9 -norm
103+
104+
# Medium regime
99105
python3 train.py --model_type compression_gan --regime med --likelihood_type logistic --warmstart -ckpt path/to/trained/model
106+
107+
# High regime
108+
python3 train.py --model_type compression_gan --regime high --warmstart -ckpt path/to/trained/model -nrb 9 -norm
109+
100110
```
101111

102112
* [`HIFIC-low`](https://drive.google.com/open?id=1hfFTkZbs_VOBmXQ-M4bYEPejrD76lAY9)

0 commit comments

Comments
 (0)