Commit
•
7b942e4
1
Parent(s):
dfa0a80
Update README.md
Browse files
README.md
CHANGED
|
@@ -211,6 +211,11 @@ metrics:
|
|
| 211 |
inference: false
|
| 212 |
---
|
| 213 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 214 |
# NLLB-MoE
|
| 215 |
|
| 216 |
This is the model card of NLLB-MoE variant.
|
|
|
|
| 211 |
inference: false
|
| 212 |
---
|
| 213 |
|
| 214 |
+
This works by upgrading bitsandbytes to the most recent version and installing this pull request of hf transformers:
|
| 215 |
+
https://github.com/huggingface/transformers/pull/26037
|
| 216 |
+
|
| 217 |
+
It uses 37 GB VRAM and is very slow, but it is properly the open access sota model in machine translation.
|
| 218 |
+
|
| 219 |
# NLLB-MoE
|
| 220 |
|
| 221 |
This is the model card of NLLB-MoE variant.
|