Remove generator-kinds paragraph from the card
Browse files
README.md
CHANGED
|
@@ -13,13 +13,6 @@ tags:
|
|
| 13 |
|
| 14 |
Checkpoints of CELL-FM.
|
| 15 |
|
| 16 |
-
The image generators come in two kinds, and the difference decides which one to use. A
|
| 17 |
-
**sequence-to-image** generator was trained with a set of proteins held out, so it images any
|
| 18 |
-
sequence, including ones it has never seen: `hpa/cellfm_seq2img.bin` and
|
| 19 |
-
`condenseq/cellfm_seq2img.bin`. A **virtual-staining** generator was trained on every protein in
|
| 20 |
-
its dataset, so it stains those proteins faithfully but is not meant for sequences outside that
|
| 21 |
-
set: `opencell/cellfm_vs.bin`, for the 1,311 OpenCell proteins.
|
| 22 |
-
|
| 23 |
| File | Model |
|
| 24 |
|---|---|
|
| 25 |
| `condenseq/cellfm_seq2img.bin` | CELL-FM CS sequence-to-image generator, trained with IDPs held out, so it images unseen 66-aa sequences (includes the ESM-C 600M encoder) |
|
|
|
|
| 13 |
|
| 14 |
Checkpoints of CELL-FM.
|
| 15 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
| File | Model |
|
| 17 |
|---|---|
|
| 18 |
| `condenseq/cellfm_seq2img.bin` | CELL-FM CS sequence-to-image generator, trained with IDPs held out, so it images unseen 66-aa sequences (includes the ESM-C 600M encoder) |
|