Commit
•
baf1ac8
1
Parent(s):
b7e1543
awq versions
Browse files
README.md
CHANGED
|
@@ -237,7 +237,7 @@ Flash attention 2 support is available both for `idefics2-8b-base` and `idefics2
|
|
| 237 |
|
| 238 |
</details>
|
| 239 |
|
| 240 |
-
|
| 241 |
|
| 242 |
<details><summary>Click to expand.</summary>
|
| 243 |
|
|
@@ -266,7 +266,7 @@ model = AutoModelForVision2Seq.from_pretrained(
|
|
| 266 |
).to(DEVICE)
|
| 267 |
```
|
| 268 |
|
| 269 |
-
</details>
|
| 270 |
|
| 271 |
# Bias, Risks, and Limitations
|
| 272 |
|
|
|
|
| 237 |
|
| 238 |
</details>
|
| 239 |
|
| 240 |
+
**4 bit quantization and module fusing**
|
| 241 |
|
| 242 |
<details><summary>Click to expand.</summary>
|
| 243 |
|
|
|
|
| 266 |
).to(DEVICE)
|
| 267 |
```
|
| 268 |
|
| 269 |
+
</details>
|
| 270 |
|
| 271 |
# Bias, Risks, and Limitations
|
| 272 |
|