Card: note the integration files are on the kw-preview branch too
Browse files
README.md
CHANGED
|
@@ -339,11 +339,13 @@ vecs = model.encode(
|
|
| 339 |
### Notes
|
| 340 |
|
| 341 |
- Revisions: the integration files are on `main`, which is what loads when no
|
| 342 |
-
`revision=` is given
|
| 343 |
-
the
|
| 344 |
-
`
|
| 345 |
-
|
| 346 |
-
|
|
|
|
|
|
|
| 347 |
- Each input item is a single modality; embed each modality in its own `encode` call
|
| 348 |
(the shared space makes the vectors directly comparable). Fused multi-modality items
|
| 349 |
(for example `{"text": ..., "audio": ...}` in one dict) are not part of the released
|
|
|
|
| 339 |
### Notes
|
| 340 |
|
| 341 |
- Revisions: the integration files are on `main`, which is what loads when no
|
| 342 |
+
`revision=` is given, and on the `v0.3-kw-preview` branch, so both the flagship and
|
| 343 |
+
the keyword-tuned model load through Sentence Transformers. The immutable release
|
| 344 |
+
tags, `v0.3-preview` among them, predate the integration and do not carry these
|
| 345 |
+
files, so the `revision=` pin shown in the `inference.py` example above does not
|
| 346 |
+
apply here. `main` carries the same flagship checkpoint as `v0.3-preview`. For a
|
| 347 |
+
reproducible load, pin the commit hash of a revision that carries the integration
|
| 348 |
+
files.
|
| 349 |
- Each input item is a single modality; embed each modality in its own `encode` call
|
| 350 |
(the shared space makes the vectors directly comparable). Fused multi-modality items
|
| 351 |
(for example `{"text": ..., "audio": ...}` in one dict) are not part of the released
|