| --- |
| license: wtfpl |
| --- |
| |
| Doesn't always work, but gets you a lot closer quickly compared to just trying keywords. |
|
|
| Dataset of 18 images, 3 vectors, batch size of 3, 3 steps gradient accumulation, trained for 500 steps or 250 epochs. |
|
|
|
|
|  |
|
|
|  |
|
|
|
|
|  |
|
|
|
|
|
|
|
|
|  |
|
|